Advertisement

A review of dose calculation approaches with cone beam CT in photon and proton therapy

      Highlights

      • Several methods to perform CBCT dose calculation are present in the literature.
      • Published CBCT dose calculation techniques are critically assessed.
      • A set of reporting standards for future CBCT dosimetric studies is recommended.

      Abstract

      Background and purpose

      The use of cone beam computed tomography (CBCT) for performing dose calculations in radiation therapy has been widely investigated as it could provide a quantitative analysis of the dosimetric impact of changes in patients during the treatment. The aim of this review was to classify different techniques adopted to perform CBCT dose calculation and to report their dosimetric accuracy with respect to the metrics used.

      Methods and materials

      A literature search was carried out in PubMed and ScienceDirect databases, based upon the following keywords: “cone beam computed tomography”, “CBCT”, “cone beam CT”, “dose calculation”, “accuracy”. Sixty-nine peer-reviewed relevant articles were included in this review: thirty-one patient studies, fifteen phantom studies and twenty-three patient & phantom studies. Most studies were found to have focused on head and neck, lung and prostate cancers.

      Results

      The techniques adopted to perform CBCT dose calculation have been grouped in six categories labelled as (1) pCT calibration, (2) CBCT calibration, (3) HU override, (4) Deformable image registration, (5) Dose deformation, and (6) Combined techniques. Differences between CBCT dose and reference dose were reported both for target volumes and OARs.

      Conclusions

      A comparison among the available techniques for CBCT dose calculations is challenging as many variables are involved. Therefore, a set of reporting standards is recommended to enable meaningful comparisons among different studies. The accuracy of the results was strongly dependent on the image quality, regardless of the methods used, highlighting the need for dose validation and quality assurance standards.

      Graphical abstract

      Keywords

      1. Introduction

      Cone beam computed tomography (CBCT) is a volumetric imaging modality currently used for verification of patient positioning in image guided radiation therapy (IGRT), and for assessing the potential need for adaptive radiotherapy (ART) [
      • Shaw Chris C.
      Cone Beam Computed Tomography (Imaging in Medical Diagnosis and Therapy).
      ,
      • Jaffray D.A.
      • Siewerdsen J.H.
      Cone-beam computed tomography with a flat-panel imager: Initial performance characterization.
      ]. In the patient treatment chain, kV systems mounted on linear accelerators (linacs) acquire images of the patient in the treatment position to verify the patient setup and hence enable adequate irradiation of the target. Geometrical variations of target volumes, organs at risk (OARs), weight changes (loss/gain), tumour regression and/or progression, which may occur during the overall treatment timeline, have resulted in CBCT images becoming an essential part of the process to ensure the calculated planned dose is delivered to the target and OARs. In this process, clinical staff investigate differences between the original planning CT (pCT) and CBCT acquired immediately prior to the treatment and use this information to assess whether to reposition the patient, replan the treatment or to proceed as scheduled with the treatment.
      Having these CBCT images available has made the prospect of also using them for dose calculations purposes, and dose accumulation during the entire treatment course, an attractive possibility. The potential for dose reporting and monitoring, particularly in a context of dose-guided ART, to quantify the dosimetric impact of significant changes in patients’ anatomy over the course of the treatment, has been investigated [
      • Ding G.X.
      • Duggan D.M.
      • Coffey C.W.
      Accurate patient dosimetry of kilovoltage cone-beam CT in radiation therapy.
      ,
      • Jaffray D.A.
      • Siewerdsen J.H.
      • Wong J.W.
      • Martinez A.A.
      Flat-panel cone-beam computed tomography for image-guided radiation therapy.
      ]. A significant number of publications have emerged on this topic with many different approaches investigated. However, the results are highly variable due to different devices, protocols, datasets, metrics of comparison, treatment sites and treatment planning systems (TPS) being reported in the different studies.
      This wide variety in the dose reconstruction methodologies and reporting methods used in this expanding field has identified a need to review the current status of the field. This work considers the current published studies which evaluate the accuracy of dosimetric calculations based on CBCT images. The advantages and disadvantages of the reported techniques are critically assessed with respect to metrics used for dosimetric analysis, for assessing feasibility and reproducibility with the aim of developing a set of recommendations for reporting future CBCT dosimetric studies.

      1.1 Images acquisition: kV CBCT, MVCT and MV CBCT in-room imaging

      Developed in the 1990′s, the first CBCT scanners for dentomaxillofacial radiology were commercially available in 2001 (NewTomQR DVT 9000; Quantitative Radiology, Verona, Italy). In the mid 2000′s CBCT started to be utilised in radiation therapy when Jaffray et al. [
      • Jaffray D.A.
      • Siewerdsen J.H.
      Cone-beam computed tomography with a flat-panel imager: Initial performance characterization.
      ] began investigating the integration of a kV x-ray source and a flat panel detector on a linac. At present, the gantry-mounted cone beam devices available are: Varian On Board Imager (OBI) (Varian Medical System, Palo Alto, USA); Elekta XVI (Elekta AB, Stockholm, Sweden); Siemens (Siemens AG, Erlangen, Germany); and Vero (BrainLAB AG, Feldkirchen, Germany & MHI, Mitsubishi Heavy Industries, Tokyo, Japan).
      In 2014 the first kV CBCT was performed on a patient undergoing proton therapy at Penn Medicine's Roberts Proton Therapy Centre (USA) [
      • Landry G.
      • Hua C.
      Current state and future applications of radiological image guidance for particle therapy.
      ]. At present, the CBCT systems available in particle therapy can be gantry, nozzle, couch, or ceiling mounted. Proteus®ONE (IBA, Louvain-la-Neuve, Belgium), ProBeam® 360° Proton Therapy System (Varian Medical System, Palo Alto, USA), Sumitomo Proton Therapy System (Sumitomo Heavy Industries, Tokyo, Japan) and PROBEAT-RT Compact Single-Room Proton Beam Therapy System (Hitachi, Ltd., Tokyo, Japan) have gantry-mounted CBCT devices. Proteus®PLUS (IBA, Louvain-la-Neuve, Belgium) has a nozzle-mounted device and ImagingRing (MedPhoton GmbH, Salzburg, Austria) uses a couch-mounted CBCT device. In-house and commercial robotic C-arm CBCT imaging devices (Hitachi, Ltd., Tokyo, Japan) have also been developed for partial gantries and fixed beam ports (e.g. ceiling).
      Megavoltage CBCT (MV CBCT) presents an alternative to the use of a kV x-ray source does not require additional equipment attached to the standard linac design. While kV imaging provides better image quality and soft tissue contrast, MV imaging is less affected by prostheses artefacts and may be better for bulky patients [
      • Broderick M.
      • McJury G.
      • Leech M.
      • Coffey M.
      • Appleyard R.
      A comparison of kilovoltage and megavoltage cone beam CT in radiotherapy.
      ,
      • Goyal S.
      • Kataria T.
      Image Guidance in Radiation Therapy: Techniques and Applications.
      ]. Currently, the MV CBCT systems in use on standard linac designs are the Artiste MVision™ (Siemens AG, Erlangen, Germany) and the In-Line kView™ (Siemens AG, Erlangen, Germany). Other linac designs include the Varian Halcyon (Varian Medical System, Palo Alto, USA), and the systems in use or available from Accuray; the TomoTherapy and the Radixact Treatment Delivery Systems (Accuray, Sunnyvale, CA), which combine the features of a linac and a helical fan beam MVCT [
      • Kraus K.M.
      • Kampfer S.
      • Wilkens J.J.
      • Schüttrumpf L.
      • Combs S.E.
      Helical tomotherapy: Comparison of Hi-ART and Radixact clinical patient treatments at the Technical University of Munich.
      ]. Although the Tomotherapy unit is not a CBCT machine, four studies performing dose calculation based on TomoTherapy MVCT were found in literature and were included (separately) in this review (Table 7). An overview of the specifics of the CBCT available devices and acquisition protocols are presented in [
      • Landry G.
      • Hua C.
      Current state and future applications of radiological image guidance for particle therapy.
      ,
      • Srinivasan K.
      • Mohammadi M.
      • Shepherd J.
      Cone Beam Computed Tomography for Adaptive Radiotherapy Treatment Planning.
      ] and summarised in Appendix A1.

      1.2 Image artefacts

      Regardless of which imaging device/modality is used, image artefacts are generated by discrepancies between the experimental setup and the mathematical assumptions established in the chosen image reconstruction algorithm. The most widely used reconstruction algorithm for CBCT is the Feldkamp algorithm [

      L.a. Feldkamp, Practical cone-beam algorithm Sfrdr I _ f, America (NY)., 1(6), pp. 612–619, 1984.

      ], either in its original form or modified [
      • Sharp G.C.
      • Kandasamy N.
      • Singh H.
      • Folkert M.
      GPU-based streaming architectures for fast cone-beam CT image reconstruction and demons deformable registration.
      ]. From a thematical point of view, this algorithm considers a 3D adaptation of the filtered back projection (FBP) method used in fan beam CT [
      • Schulze R.
      • et al.
      Artefacts in CBCT: A review.
      ]. When using FBP, a dependency between the distance from the central plane and the image quality degradation, has been shown [
      • Kalender W.A.
      • Kyriakou Y.
      Flat-detector computed tomography (FD-CT).
      ]. Consequently, when performing CBCTs on large FOVs, the larger amount of detected scatter radiation may affect the image quality. A valid strategy currently adopted to reduce the scatter contribution considers the use of a beam shaping device, such as bowtie filters and collimators, or by placing anti-scatter x-ray grid on the detector, which decrease the primary dose but also reduces the cupping artefacts. An overview of the main CBCT artefacts, including their appearances and causes, is presented in the supplementary material (Table S1) while an extensive description can be found in [
      • Srinivasan K.
      • Mohammadi M.
      • Shepherd J.
      Cone Beam Computed Tomography for Adaptive Radiotherapy Treatment Planning.
      ].
      Image artefacts may have a non-negligible impact on any dose calculation since treatment planning relies on the accurate determination of the CT numbers, which corresponds to the photons or protons attenuation inside the scanned object. Therefore, corrections techniques may be implemented and different protocols can be selected when acquiring CBCT images in order to ensure that the image artefacts are minimised. Post-processing algorithms can be used to correct for CBCT imaging artefacts such as scattering, beam hardening, ring artefacts, motion artefacts, and cupping artefacts [
      • Srinivasan K.
      • Mohammadi M.
      • Shepherd J.
      Cone Beam Computed Tomography for Adaptive Radiotherapy Treatment Planning.
      ,
      • Schulze R.
      • et al.
      Artefacts in CBCT: A review.
      ].
      Finally, the implementation of iterative reconstruction algorithms [
      • Mao W.
      • et al.
      Evaluation and clinical application of a commercially available iterative reconstruction algorithm for CBCT-based IGRT.
      ,
      • Gardner S.J.
      • et al.
      Improvements in CBCT Image Quality Using a Novel Iterative Reconstruction Algorithm: A Clinical Evaluation.
      ] and the introduction of machine learning techniques into the tomographic image reconstruction process [
      • Kida S.
      • et al.
      Cone Beam Computed Tomography Image Quality Improvement Using a Deep Convolutional Neural Network.
      ] have more recently started being investigated to improve the image quality relative to FBP.

      1.3 Treatment planning systems and dose calculation algorithms

      CBCT dose calculations are typically carried out using a standard TPS. There are many different types of TPS available which may have dose calculation algorithms each of which have limitations. These limitations can depend on how well the physical interactions are modelled within the dose calculation algorithm and how well the beam modelling is carried out when commissioning the TPS [

      J. B. Smilowitz et al., AAPM Medical Physics Practice Guideline 5.a.: Commissioning and QA of Treatment Planning Dose Calculations – Megavoltage Photon and Electron Beams, J Appl Clin Med Phys, 17(1), 2016, p. 6166, doi: 10.1120/jacmp.v17i1.6166.

      ]. Knöös et al. [
      • Knöös T.
      • et al.
      Comparison of dose calculation algorithms for treatment planning in external photon beam therapy for clinical situations.
      ] grouped the model-based algorithms for TPS in external photon beam therapy into three types: Type A and B, which differ on the level of modelling of the lateral electron transport and Type C, the last generation of dose calculation algorithms, solving the electron transport either using the Boltzmann equations or Monte Carlo (MC) calculations.
      Type A and B algorithms have shown to have dose <1.0% of expected within the target in prostate [
      • Knöös T.
      • et al.
      Comparison of dose calculation algorithms for treatment planning in external photon beam therapy for clinical situations.
      ,
      • Bufacchi A.
      • Nardiello B.
      • Capparella R.
      • Begnozzi L.
      Clinical implications in the use of the PBC algorithm versus the AAA by comparison of different NTCP models/parameters.
      ], although a Type C algorithm is superior if air gaps are within the rectum [
      • Kim K.-H.
      • et al.
      Dosimetric and radiobiological comparison in different dose calculation grid sizes between Acuros XB and anisotropic analytical algorithm for prostate VMAT.
      ]. In lung, improved dosimetric accuracy has been shown by employing Type B or Type C algorithms. Similar results have been observed for head and neck patients [
      • Vanderstraeten B.
      • et al.
      Accuracy of patient dose calculation for lung IMRT: A comparison of Monte Carlo, convolution/superposition, and pencil beam computations.
      ,
      • Tsuruta Y.
      • et al.
      Dosimetric comparison of Acuros XB, AAA, and XVMC in stereotactic body radiotherapy for lung cancer.
      ]. In proton therapy, MC has been shown to better model heterogeneities compared to the standard pencil beam model [
      • Saini J.
      • et al.
      Advanced Proton Beam Dosimetry Part I: review and performance evaluation of dose calculation algorithms.
      ]. The TPS with associated dose calculation algorithms used in the identified publications included in this review are listed in the supplementary material (Table S2).
      The choice of an appropriate algorithm is crucial to avoid inaccuracies in the dose calculation. However, when comparing the dose calculated between different studies, the beam model used should be taken into account in order to avoid biased comparisons, since beam parameters, tumour size and location may all have a non-negligible impact on the dose calculation accuracy depending on the beam model used [
      • Dufreneix S.
      • et al.
      Design of experiments in medical physics: Application to the AAA beam model validation.
      ]. Guidelines to follow for the commissioning of a new TPS are provided in [
      • Ma C.M.C.
      • et al.
      Beam modeling and beam model commissioning for Monte Carlo dose calculation-based radiation therapy treatment planning: Report of AAPM Task Group 157.
      ], but in order to have a standard beam model for dose calculation (i.e. an unbiased comparison), beam data should be provided. The different contributions to the dose prediction errors from the beam model and the dose calculation algorithm are not separated and the evaluation of this is outside the scope of this review.

      1.4 Description of the radiotherapy workflow

      Radiotherapy treatment plans are typically created using pCT images, in which target volumes (gross tumour volume, GTV – clinical target volume, CTV – internal target volume, ITV – planning target volume, PTV) and OARs are contoured and plans created. A calibration curve is then used to convert the CT numbers (measured in Hounsfield unit – HU) into electron density (ED), relative electron density (RED), physical density (PD) or stopping power ratio (SPR) so that the attenuation in the object can be calculated. These curves are either provided within the TPS and verified during commissioning (denoted as the “TPS calibration curve”) or created by the user during commissioning (denoted as the “user specified calibration curve”). Calibration curves converting CBCT HU values into PD/RED/ED/SPR can also be generated using phantom/patient/population-specific measurements as described in Section 3.1.2.
      Before delivering the treatment, CBCT scans are typically acquired and registered (via rigid or deformable registration) onto the pCT, to verify the positioning of the patient. In case of anatomical differences between pCT and CBCT, physicists, radiographers and clinicians decide if the original plan can still be delivered or if replanning the patient is necessary, resulting in a new rescan CT (rCT), new contours, and new plan. In this scenario, the possibility of performing a dose calculation based on CBCT images would provide a quantitative analysis of the dosimetric impact of any changes, and could be work and time saving for both clinical staff and patients. Avoidance of rescanning will also reduce the dose received by the patient.

      2. Methods

      2.1 Search criteria

      PubMed and ScienceDirect databases were used to identify published articles reporting on the accuracy of dose calculations performed on CBCT images as compared to the chosen reference dose. Using a combination of the keywords “cone beam computed tomography”, “CBCT”, “cone beam CT”, “dose”, “calculation”, “accuracy” over one hundred peer-reviewed articles published in English between January 2006 and December 2019 were retrieved. Articles about dental CBCT, imaging dose due to CBCT, non-radiotherapy topics or those not presenting a quantitative analysis were excluded, resulting in sixty-nine peer-reviewed articles being included.
      A further four peer-reviewed articles were added after including the words “tomotherapy” and “MV CT” to the keywords. Studies were excluded from the review that investigated the impact of geometric shifts or anatomical changes in on the dose calculation, reported on the additional dose delivered to the patients from the daily/weekly CBCT image acquisition or were focused on CBCT image quality, rather than presenting a method to perform dose calculations based on CBCT. Including the words “gamma knife”, “electrons”, or “heavy ion” in the search criteria identified publications in PubMed and ScienceDirect, but none of these were found to specifically investigate CBCT dose calculation methods.

      2.2 Reference dose

      To assess the feasibility of performing dose calculation based on CBCT, in the majority of the cases, the dose calculated on pCT or rCT (acquired on a date closer to the treatment start or during the treatment itself to monitor the response of the patient) were considered the ground truth, i.e. the reference dose to be used to evaluate the difference from CBCT dose calculations. Another possibility described in literature is the use of “virtual” CT (vCT), generated by deforming pCTs onto the geometry of CBCTs [
      • Thummerer A.
      • et al.
      Comparison of CBCT based synthetic CT methods suitable for proton dose calculations in adaptive proton therapy.
      ]. Both deformable image registration [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ,
      • Veiga C.
      • et al.
      First Clinical Investigation of Cone Beam Computed Tomography and Deformable Registration for Adaptive Proton Therapy for Lung Cancer.
      ,
      • Veiga C.
      • et al.
      A comprehensive evaluation of the accuracy of CBCT and deformable registration based dose calculation in lung proton therapy.
      ] and artificial intelligence-based methods [
      • Kida S.
      • et al.
      Cone Beam Computed Tomography Image Quality Improvement Using a Deep Convolutional Neural Network.
      ,
      • Hansen D.C.
      • et al.
      ScatterNet: a convolutional neural network for cone-beam CT intensity correction.
      ,
      • Landry G.
      • et al.
      Comparing Unet training with three different datasets to correct CBCT images for prostate radiotherapy dose calculations.
      ,
      • Liang X.
      • et al.
      Generating synthesized computed tomography (CT) from cone-beam computed tomography (CBCT) using CycleGAN for adaptive radiation therapy.
      ] were adopted for generating vCT in dose calculation studies. Only two studies reported on the use of farmer chambers and Gafchromic films to measure the actual ground truth dose [
      • Hatton J.
      • McCurdy B.
      • Greer P.B.
      Cone beam computerized tomography: The effect of calibration of the Hounsfield unit number to electron density on dose calculation accuracy for adaptive radiation therapy.
      ,
      • Yohannes I.
      • Prasetio H.
      • Kallis K.
      • Bert C.
      Dosimetric accuracy of the cone-beam CT-based treatment planning of the Vero system: a phantom study.
      ] and the measured dose was compared both with the pCT and the CBCT calculated dose.

      2.3 Metrics

      Since many different metrics were used, we have chosen to report only the metrics most commonly used in the literature (summarised in Appendix A.2) in order to facilitate the comparison among different studies, e.g. gamma analysis, mean/median PTV/GTV/PTV dose difference, and dose covering specific percentages of the PTV/CTV/GTV.

      3. Results

      3.1 CBCT dose calculation techniques

      Sixty-nine studies were found highlighting the potential of using CBCT images for dose calculation purposes: thirty-one patient studies, fifteen phantom studies and twenty-three patient and phantom studies. The techniques adopted in these studies have been grouped in the following categories:
      • 1-
        pCT calibration: The dose is calculated on the CBCT images using the pCT calibration curve adopted for the pCT images. The patient and phantom studies retrieved from the literature adopting this method are listed in Table 1a, Table 1b, respectively.
        Table 1aCBCT dose calculation studies using the pCT calibration for different tumour sites in patients, in chronological order.
        AuthorMake / ModelTPSAlgorithmSiteProtocolArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Ding et al. (2007)
        • Ding G.X.
        • et al.
        A study on adaptive IMRT treatment planning using kV cone-beam CT.
        Varian TrilogyEclipseAAAProstate (2)Not defined (HF, 125kV, 80mA)Inhomogeneity (modified Batho)n/spCTDmean-PTV0.5n/sDmean: Bld = 1.3, Rct = 0.6, SemVsc = 0.6, Urt = 0.7, LHip = 0.1, RHip = 0.0
        H&N (1)Not defined (FF, 125kV, 80mA)0.4Dmean: LPar = 0.5, RPar = 0.5, SC = 1.2
        Richter et al. (2008)
        • Richter A.
        • et al.
        Investigation of the usability of conebeam CT data sets for dose calculation.
        Elekta Synergy XVIPinnacle ADAC (v7.6s)n/sProstate (5)Pelvisn/aUser specified CT# to PDpCTDmean19.1 ± 3.4n/sDVH of one patient per site
        Thorax (5)Thorax13.8 ± 10.5
        Head (5)Head16.2 ± 12.7
        Lee et al. (2008)
        • Lee L.
        • Le Q.T.
        • Xing L.
        Retrospective IMRT Dose Reconstruction Based on Cone-Beam CT and MLC Log-File.
        Varian TrilogyEclipseAAAH&N (6)n/sn/aUser specified CT# to REDpCTDPTV-95%<3.0n/sDmax: SC = 1.0, BS < 1.1
        Boggula et al. (2009)
        • Boggula R.
        • et al.
        A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
        Elekta Synergy XVICorvusPBProstate (3)Pelvisn/aUser specified CT# to REDpCTAbsolute33.3 ± 18.83%/3mm33.6 ± 23.6n/s
        Onozato et al. (2014)
        • Onozato Y.
        • et al.
        Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
        Varian OBI (v1.5)Eclipse (v8.6.15)AAAProstate (10)Pelvisn/aUser specified CT# to EDpCTDmean-PTV1.2 ± 0.52%/2mm97.0Dmean: Bld = 1.24 ± 0.91, Rct = 1.02 ± 1.11
        Noufal et al. (2016)
        • Noufal M.P.
        • Abdullah K.K.
        • Niyas P.
        • Sankaran T.S.
        • Sasindaran P.R.
        Analysis of dosimetric impacts of cone beam computed tomography-based volumetric modulated arc therapy planning.
        Varian Clinac IX OBIEclipseAAAProstate (10)Chestn/aUser specified CT# to REDpCTDPTV-95%1.4 ± 0.92%/2mm91.0 ± 0.0Dmean: Bld < 1.5, Rct < 2.3
        Brain (10)Head0.9 ± 0.688.8 ± 0.1Dmean: RLens < 1.5, LLens < 1.5, ROptNrv < 1.5, LOptNrv < 2.0, Ch < 1.0, BS < 2.0
        H&N (7)2.9 ± 3.279.0 ± 6.0Dmean: RPar < 1.5, LPar < 1.0, Mnd < 1.0, SC < 2.0
        De Smet et al. (2016)
        • De Smet M.
        • Schuring D.
        • Nijsten S.
        • Verhaegen F.
        Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
        Varian Truebeam STx (v1.5)EclipseAXB 10.0Lung (6)Thoraxn/aUser specified CT# to PDrCTDmean-PTV<1.0n/sHistogram – Lng-GTV(V20Gy) < 2.0, Dmax(0.1cc): SC < 2.0, Oes(V35Gy) < 3.0 – Dmean: Lng < 2.5, H < 2.5, Oes < 3.0
        Elekta Synergy XVI (v5.0)Pinnacle (v9.8)n/sLung (10)Clinical optimizedn/an/s<6.0Histogram – Lng-GTV(V20Gy) < 8.0, Dmax(0.1cc): SC < 8.0, Oes(V35Gy) < 8.0 – Dmean: Lng < 6.0, H < 8.0, Oes < 9.0
        Arai et al. (2017)
        • Arai K.
        • et al.
        Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
        Varian Clinac iX OBI 1.6XiO-MPBH&N (10)Not defined (FF, 100kV, 10mA)n/an/spCTDPTV-98%27.1 ± 31.83%/3mm87.8 ± 7.4Dmean: LPar = 5.42 ± 3.06
        Giacometti et al. (2018)
        • Giacometti V.
        • et al.
        An evaluation of techniques for dose calculation on cone beam computed tomography.
        Varian TruebeamEclipse (v13.5)AAAProstate (5)Pelvisn/aUser specified CT# to REDpCTDmedianPTV-99%0.02%/0.1mm

        (50% threshold)
        99.8Dmedian: Bld = 0.0, Rct = 0.4
        Lung (5)Thorax−0.696.0Dmedian: SC = −0.2, BS = −0.1
        H&N (5)Head−0.495.0Dmax(0.01cc): SC = 0.2, Dmax(0.1cc): Oes = 0.1, H = 0.5
        Marchant et al. (2018)

        Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

        Elekta Synergy XVI (v4.5-5.0)Pinnacle (v9.8)n/sPelvis (15)n/sn/an/spCTDmean-PTV7.5 ± 2.6n/sRct(V40Gy) = 9.62 ± 6.93
        Shading0.1 ± 0.3Rct(V40Gy) = 0.09 ± 0.55
        Lung (15)n/a3.9 ± 2.1Dmax(1cc): SC = 4.11 ± 2.59
        Shading0.4 ± 0.7Dmax(1cc): SC = 0.35 ± 0.74
        H&N (14)n/a0.8 ± 0.3Dmean: CPar = 0.15 ± 0.58
        Shading0.0 ± 0.2Dmean: CPar = 0.10 ± 0.34
        Wang et al. (2019)
        • Wang T.
        • et al.
        Dosimetric study on learning-based cone-beam CT correction in adaptive radiation therapy.
        Varian TruebeamEclipse (v13.6)AAAProstate (10)n/sn/an/spCTDmean-PTV0.8 ± 0.51%/1mm94.5Dmean: Bld = 0.5 ± 0.3, Rct = 0.5 ± 0.6, Fem = 0.2 ± 0.3
        Scatter (machine learning
        • Lei Y.
        • et al.
        Improving Image Quality of Cone-Beam CT Using Alternating Regression Forest.
        )
        0.3 ± 0.199.0Dmean: Bld = 0.2 ± 0.1, Rct = 0.2 ± 0.3, Fem = 0.1 ± 0.3
        Varian Truebeam (Novalis Tx)Brain (12)n/a0.2 ± 0.1(99.0; 100.0)Dmean: BS = 0.02 ± 0.03, OptNrv = 0.02 ± 0.05, Ch = 0.02 ± 0.05
        Scatter (machine learning
        • Lei Y.
        • et al.
        Improving Image Quality of Cone-Beam CT Using Alternating Regression Forest.
        )
        0.1 ± 0.1(99.0; 100.0)Dmean: BS = 0.00 ± 0.00, OptNrv = 0.01 ± 0.02, Ch = 0.00 ± 0.00
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; AXB = Acuros XB; H&N = head and neck; HF = half fan; FF = full fan; CT# = CT number; ED = electron density; RED = relative electron density; PD = physical density; pCT = planning CT; rCT = rescan CT; CTV = clinical target volume; GTV = gross tumour volume; PTV = planning target volume; Dmax = maximum dose; Dmean = mean dose; Dmedian = median dose; DPTV-#% = dose delivered to #% of PTV; Dmedian-PTV-#% = median dose delivered to #% of PTV; Dmax(#cc) = maximum dose to #cm3 of the PTV; R = right; L = left; C = contra; Bld = bladder; Rct = rectum; SemVsc = seminal vescicle; Urt = urethra; Par = parotid; SC = spinal cord; OptNrv = optic nerve; Ch = chiasm; BS = brain stem; Mnd = mandible; Lng = lung; Oes = oesophagus; H = heart; Fem = femur; V(#Gy) = volume receiving #Gy or more; n/s = not supplied; n/a = not applicable.
        Table 1bCBCT dose calculation studies using the pCT calibration for different phantoms, in chronological order.
        AuthorMake / ModelTPSAlgorithmPhantomProtocolArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Yoo et al. (2006)
        • Yoo S.
        • Yin F.F.
        Dosimetric feasibility of cone-beam CT-based treatment planning compared to CT-based treatment planning.
        Varian OBIEclipsen/sCIRS 002HAS6N small homogeneous cylindrical phantom (head)Not defined (125kVp, 80mA)Inhomogeneity (modified Batho)n/spCTDmean-PHANTOM(−0.6; −0.3)n/sn/s
        CIRS 002HAS6N large homogeneous ellipsoidal phantom (body)(−2.5; 0.7)
        CIRS 008 dynamic thorax (inhomogeneous phantom)(−3.4; −0.3)
        Lee et al. (2008)
        • Lee L.
        • Le Q.T.
        • Xing L.
        Retrospective IMRT Dose Reconstruction Based on Cone-Beam CT and MLC Log-File.
        Varian TrilogyEclipseAAACatphan 600n/sn/aUser specified CT# to REDpCTDmax<0.5n/sn/s
        Richter et al. (2008)
        • Richter A.
        • et al.
        Investigation of the usability of conebeam CT data sets for dose calculation.
        Elekta Synergy XVIPinnacle ADAC (v7.6s)n/sCatphan (CTP503)Thorax/Pelvisn/apCTDmean-PHANTOM(6.8 ± 5.6; 18.4 ± 3.7)n/sn/s
        Head(8.8 ± 6.3; 19.9 ± 2.5)
        Aubry et al. (2009)
        • Aubry J.F.
        • Cheung J.
        • Morin O.
        • Gottschalk A.
        • Beaulieu L.
        • Pouliot J.
        Correction of megavoltage cone-beam CT images of the pelvic region based on phantom measurements for dose calculation purposes.
        Siemens MVisionn/sn/sRANDO pelvisn/sn/an/spCTDmean-PHANTOM(−4.0; 8.0)n/sn/s
        Cupping & missing data(−3.0; 3.0)
        Boggula et al. (2009)
        • Boggula R.
        • et al.
        A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
        Elekta Synergy XVICorvus (v6.2)PBPelvic phantomPelvisn/aUser specified CT# to REDpCTn/s3%/3mm68.9 ± 5.5n/s
        Guan et al. (2009)
        • Guan H.
        • Dong H.
        Dose calculation accuracy using cone-beam CT (CBCT) for pelvic adaptive radiotherapy.
        Varian OBI (v1.4)EclipseAAAPelvic phantomPelvisn/aUser specified CT# to REDpCTDmean-PTV(−1.1; −0.2)n/sDmean: Bld = (−1.1; −0.4), Rct = (−0.2; 0.1)
        Hatton et al (2009)
        • Hatton J.
        • McCurdy B.
        • Greer P.B.
        Cone beam computerized tomography: The effect of calibration of the Hounsfield unit number to electron density on dose calculation accuracy for adaptive radiation therapy.
        Varian 2100 IX OBIEclipse (v8.6)AAASlabs phantomPelvisn/aUser specified CT# to REDIonization chamberDmean-SLABS<5.0n/sn/s
        RT01 trial semi-anthropomorphic phantomDmean-PTV<1.0
        Qian et al. (2010)
        • Qian J.
        • et al.
        Dose reconstruction for volumetric modulated arc therapy (VMAT) using cone-beam CT and dynamic log files.
        Varian TrilogyEclipseAAACatphan 600Not defined (HF, 125kVp, 80mA, 25ms)n/aUser specified CT# to REDpCTDmean-PTV0.7n/sn/s
        Hu et al. (2011)
        • Hu C.C.
        • et al.
        Practically acquired and modified cone-beam computed tomography images for accurate dose calculation in head and neck cancer.
        Elekta Synergy XVIPinnacle (v7.6c)n/sCatphan 503Not defined (120kVp, 32mA, 10ms)n/aUser specified CT# to PDpCTDmean-PHANTOM(4.2; 4.4)n/sn/s
        Wen et al. (2012)
        • Wen N.
        • et al.
        Evaluation of the deformation and corresponding dosimetric implications in prostate cancer treatment.
        Varian TrilogyEclipse (v8.6)PBRANDO pelvisPelvisn/aUser specified CT# to EDpCTDmean-PTV0.0n/sn/s
        Inhomogeneity (modified Batho)(−1.8; −1.4)n/sDmean: Bld = (−0.8; −0.7), Rct = (−0.7; 2.7)
        Onozato et al. (2014)
        • Onozato Y.
        • et al.
        Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
        Varian OBI (v1.5)Eclipse (v8.6.15)AAAPelvic phantomPelvisn/aUser specified CT# to EDpCTDmean-PTV0.4n/sHistogram – Dmean: Bld = 0.4: Rct = 0.1
        Annkah et al. (2014)
        • Annkah J.K.
        • et al.
        Assessment of the dosimetric accuracies of CATPhan 504 and CIRS 062 using kV-CBCT for performing direct calculations.
        Varian TruebeamEclipsen/sCatphan 504Not defined (HF, 125kV, 254mAs)n/aUser specified CT# to EDpCTDmean-PHANTOM_CENTRE ± 14.0n/sn/s
        CIRS062M ± 5.0
        Noufal et al. (2016)
        • Noufal M.P.
        • Abdullah K.K.
        • Niyas P.
        • Sankaran T.S.
        • Sasindaran P.R.
        Analysis of dosimetric impacts of cone beam computed tomography-based volumetric modulated arc therapy planning.
        Varian Clinac IX OBIEclipseAAACatphan 504n/sn/aUser specified CT# to REDpCTn/s2%/2mm96.0Dmean: ROptNrv = 0.8, LOptNrv = 0.2, Ch = 0.1, BS = 0.2
        Arai et al. (2017)
        • Arai K.
        • et al.
        Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
        Varian Clinac iX OBI 1.6XiO-MPBPelvic phantomPelvisn/an/spCTDPTV-98%0.63%/3mm96.5Dmean: Rct = 0.58, Bld = 0.89
        H&N phantomNot defined (FF, 100kV, 10mA)3.498.9Dmean: LPar = 0.67
        Marchant et al. (2018)

        Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

        Elekta Synergy XVI (v4.5-5.0)Pinnacle (v9.8)n/sAlderson Rando (prostate)n/sn/an/spCTDmean-PTV4.7n/sRct(V40Gy) = 24.1
        Shading−0.3Rct(V40Gy) = −1.4
        Alderson Rando (lung)n/a-1.2Dmax(1cc): SC = 4.2
        Shading−0.4Dmax(1cc): SC = −0.3
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; H&N = head and neck; HF = half fan; FF = full fan; CT# = CT number; ED = electron density; RED = relative electron density; PD = physical density; pCT = planning CT; CTV = clinical target volume; PTV = planning target volume; Dmax = maximum dose; Dmean = mean dose; Dmedian = median dose; DPTV-#% = dose delivered to #% of PTV; Dmax(#cc) = maximum dose to #cm3 of the PTV; OAR = organ at risk; R = right; L = left; Bld = bladder; Rct = rectum; OptNrv = optic nerve; Ch = chiasm; Par = parotid; BS = brain stem; SC = spinal cord; V(#Gy) = volume receiving #Gy or more; n/s = not supplied; n/a = not applicable.
      • 2-
        CBCT calibration: The dose is calculated on the CBCT images using the CBCT calibration curve generated via phantom/patient/population-specific measurements. The patient and phantom studies retrieved from the literature adopting this method are listed in Table 2a, Table 2b, respectively.
        Table 2aDose calculation studies using the CBCT calibration for different tumour sites in patients, in chronological order.
        AuthorMake/modelTPSAlgorithmSiteProtocolCBCT CalibrationArtefact CorrectionReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Morin et al. (2007)
        • Morin O.
        • et al.
        Dose calculation using megavoltage cone-beam CT.
        Siemens MVisionPinnacle (v7.6)n/sNPC (1)rFOVPhantom CBCT# to REDUniformitypCTn/s3%/3mm98.0DVH
        Tongue (1)3%/3mm

        (30% threshold)
        (78.0; 84.0)
        Richter et al. (2008)
        • Richter A.
        • et al.
        Investigation of the usability of conebeam CT data sets for dose calculation.
        Elekta Synergy XVIADAC Pinnacle (v7.6s)n/sProstate (5)PelvisCatphan(CTP503) CBCT# to PDn/apCTDmean12.7 ± 1.5n/sDVH
        Population-based CBCT# to PD1.3 ± 1.0
        Patient-specific CBCT# to PD1.2 ± 0.9
        Thorax (5)ThoraxCatphan(CTP503) CBCT# to PD6.2 ± 4.7
        Population-based CBCT# to PD1.7 ± 2.4
        Patient-specific CBCT# to PD1.7 ± 2.0
        Head (5)HeadCatphan(CTP503) CBCT# to PD1.6 ± 2.2
        Population-based CBCT# to PD1.4 ± 1.9
        Patient-specific CBCT# to PD1.4 ± 1.9
        Cheung et al. (2009)
        • Cheung J.
        • Aubry J.F.
        • Yom S.S.
        • Gottschalk A.R.
        • Celi J.C.
        • Pouliot J.
        Dose Recalculation and the Dose-Guided Radiation Therapy (DGRT) Process Using Megavoltage Cone-Beam CT.
        Siemens MVisionKonRad InversePBPelvis (2)Low dose image acquisitionPhantom CBCT# – PDScatter
        • Aubry J.F.
        • Cheung J.
        • Morin O.
        • Gottschalk A.
        • Beaulieu L.
        • Pouliot J.
        Correction of megavoltage cone-beam CT images of the pelvic region based on phantom measurements for dose calculation purposes.
        pCTn/sn/sDVH
        H&N (6)DGTV-95%0.10Dmean: LPar = 5.8, RPar = 18.0, Lrx = 6.7, D1%: SC = 2.2, Mnd = 1.8, BS = 1.1, LTMJ = −1.6, RTMJ = −0.5
        Hu et al. (2010)
        • Hu W.
        • Ye J.
        • Wang J.
        • Ma X.
        • Zhang Z.
        Use of kilovoltage X-ray volume imaging in patient dose calculation for head-and-neck and partial brain radiation therapy.
        Elekta Synergy XVIPinnacle (v8.0d)CCSCBrain (3)Not defined (S20, 100kV, 65mAs)H&N one patient-specific CBCT# to PDn/apCTn/s2%/2mm (10% threshold)(99.5; 100.0)Reported for individual patients
        H&N (3)
        Fotina et al. (2012)
        • Fotina I.
        • Hopfgartner J.
        • Stock M.
        • Steininger T.
        • Lütgendorf-Caucig C.
        • Georg D.
        Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
        Elekta Synergy XVIiPlan (v4.5.0)XV MCProstate (10)Prostate (no bowtie)Phantom (Catphan) CBCT# to EDn/aPopulation-based CBCT# to EDDmedian-PTV−2.0n/sDmean: Bld = 0.4, Rct = 0.9, FemHeads = 0.2
        Lung (10)Lung (no bowtie)Phantom (Catphan) CBCT# to ED−2.2Dmax: SC = −0.3, OES = 0.0, Dmean: Lng = 0.0
        H&N (7)Head (no bowtie)Phantom (Catphan) CBCT# to ED4.0Dmax: SC = 5.1, BS = 4.2, Dmean: Par = 1.5
        Wang et al. (2013)
        • Wang J.
        • et al.
        Using corrected cone-beam CT image for accelerated partial breast irradiation treatment dose verification: the preliminary experience.
        Elekta Synergy XVIPinnacle (v8.0)n/sBreast (10)Not defined (S20, 120kVp, 31mAs)CBCT# to PD
        • Hu W.
        • Ye J.
        • Wang J.
        • Ma X.
        • Zhang Z.
        Use of kilovoltage X-ray volume imaging in patient dose calculation for head-and-neck and partial brain radiation therapy.
        ScatterpCTDPTV-95%0.0n/sIps lng(V30) = 0.0, H(V5) = 0.0
        Held et al. (2015)
        • Held M.
        • Sneed P.K.
        • Fogh S.E.
        • Pouliot J.
        • Morina O.
        Feasibility of MV CBCT-based treatment planning for urgent radiation therapy: Dosimetric accuracy of MV CBCT-based dose calculations.
        Siemens Artiste & In-Line kView MV CBCTPinnacle (v9.2)n/sPelvis (9)eFOVPhantom (CIRS 062 + water cylinder) CBCT# to PD – eFOVn/apCTn/s3%/3mm(90.0; 99.0)n/s
        Thorax (9)eFOV(47.0; 100.0)
        Brain (7)rFOVPhantom (CIRS 062 + water cylinder) CBCT# to PD – rFOV(99.9; 100.0)
        Extremities (2)rFOV(92.0; 99.0)
        De Smet et al. (2016)
        • De Smet M.
        • Schuring D.
        • Nijsten S.
        • Verhaegen F.
        Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
        Varian Truebeam (v1.5)EclipseAXB 10.0Lung (6)ThoraxPopulation-based (thorax) CBCT# to PDn/arCTDmean-PTV<0.5n/sHistogram – Lng-GTV(V20Gy) < 1.0 – Dmean: Lng < 1.0, H < 3.0, OES < 2.0 – Dmax(0.1cc): SC < 1.0 – OES(V35Gy) < 2.0
        Patient-specific CBCT# to PD<1.0Histogram – Lng-GTV(V20Gy) < 1.0 – Dmean: Lng < 1.0, H < 3.0, OES < 1.5 – Dmax(0.1cc): SC < 1.5 – OES(V35Gy) < 2.0
        Elekta Synergy XVI (v5.0)Pinnacle (v9.8)n/sLung (10)Clinical optimizedPopulation-based (thorax) CBCT# to PD<1.5Histogram – Lng-GTV(V20Gy) < 3.0 – Dmean: Lng < 2.0, H < 3.5, OES < 2.0 – Dmax(0.1cc): SC < 2.5 – OES(V35Gy) < 2.5
        Patient-specific CBCT# to PD<1.0Histogram – Lng-GTV(V20Gy) < 1.5 – Dmean: Lng < 1.0, H < 3.0, OES < 1.5 – Dmax(0.1cc): SC < 2.5 – OES(V35Gy) < 2.0
        Kaliyaperumal et al. (2017)

        Kaliyaperuma IV, et al., Study of Variation in Dose Calculation Accuracy Between kV Cone‑Beam Computed Tomography and kV fan‑Beam Computed Tomography, J Med Phys, 42(3), 2017, p. 171–180, doi: 10.4103/jmp.JMP.

        Varian Clinac IX OBIEclipse (v10.0)AAAPelvis (1)ThoracicPhantom (Catphan CTP404) CBCT# to EDn/apCTAbsolute<1.0n/sn/s
        H&N (1)Head
        Giacometti et al. (2018)
        • Giacometti V.
        • et al.
        An evaluation of techniques for dose calculation on cone beam computed tomography.
        Varian TruebeamEclipse (v13.5)AAAProstate (5)PelvisPhantom (Inner cylinder CIRS062M + wax + water blocks) CBCT# to REDn/apCTDmedianPTV-99%−0.22%/0.1mm (50% threshold)99.6Dmedian: Bld = 0.0, Rct = −0.1
        Lung (5)ThoraxPhantom (Outer ring CIRS062M + inner cylinder low density material + wax + water blocks) CBCT# to RED−0.799.6Dmedian: SC = 0.0, BS = 0.2
        H&N (5)HeadPhantom (CIRS062M + wax + water blocks) CBCT# to RED−0.194.6Dmax(0.01cc): SC = 0.0, Dmax(0.1cc): Oes = 0.0, H = 0.0
        Kaplan et al. (2018)
        • Kaplan L.P.
        • Elstrøm U.V.
        • Møller D.S.
        • Hoffmann L.
        Cone beam CT based dose calculation in the thorax region.
        Varian Clinac IX OBIEclipse (v13.7.14)AAALung (12)Low dose thoraxPhantom (CIRS062M) CBCT# to REDn/asCTDmedian (DVH)<2.0n/sBox Plot – Lng(V20Gy)∼0.0, H(V25Gy) < 2.0, Dmax(0.01cc): Brnc < 2.0, SC < 1.0, Oes < 1.0
        Jarema et al. (2019)
        • Jarema T.
        • Aland T.
        Using the iterative kV CBCT reconstruction on the Varian Halcyon linear accelerator for radiation therapy planning for pelvis patients.
        Varian HalcyonEclipse (v15.6.03)AAAPelvis (10)PelvisPhantom (CIRS062M) CBCT# to EDn/apCTDmean-PTV(−1.7; 3.0)1%/1mm (10% threshold)94.4 ± 6.1Dmean: Bld =  (−1.3; 7.9), Rct = (−2.6; 28.1)
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; CCSC = collapsed cone superposition convolution; AXB = Acuros XB; MC = Monte Carlo; H&N = head and neck; eFOV = extended field of view; rFOV = regular field of view; CT# = CT number; CBCT# = CBCT value; ED = electron density; RED = relative electron density; PD = physical density; pCT = planning CT; rCT = rescan CT; GTV = gross tumour volume; PTV = planning target volume; Dmax = maximum dose; Dmean = mean dose; Dmedian = median dose; DPTV-#% = dose delivered to #% of PTV; DGTV-#% = dose delivered to #% of GTV; Dmedian-PTV-#% = median dose delivered to #% of PTV; Dmax(#cc) = maximum dose to #cm3 of the PTV; D#% =  minimum doses received by at least #% of the volume that received the largest dose; OAR = organ at risk; R = right; L = left; Bld = bladder; Rct = rectum; Par = parotid; SC = spinal cord; BS = brain stem; Mnd = mandible; Lng = lung; Oes = oesophagus; H = hearth; TMJ = temporomandibular joint; Fem = femur; Lrx = larynx; Brnc = bronchi; Ips = ipsilateral; V(#Gy) = volume receiving #Gy or more; n/s = not supplied; n/a = not applicable.
        Table 2bDose calculation studies on several phantoms using the CBCT calibration obtained with different phantoms configurations, in chronological order.
        AuthorMake/modelTPSAlgorithmPhantomProtocolCBCT CalibrationArtefact CorrectionReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Morin et al. (2007)
        • Morin O.
        • et al.
        Dose calculation using megavoltage cone-beam CT.
        Siemens MVisionPinnacle (v7.6)n/sWater cylinderrFOVWater cylinder+inserts CBCT# to REDn/apCTDmean-PHANTOM<5.0n/sn/s
        n/sUniformity<1.0
        Yang et al. (2007)
        • Yang Y.
        • Schreibmann E.
        • Li T.
        • Wang C.
        • Xing L.
        Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation.
        Varian TrilogyEclipsePBCatphan 600 (CTP404)PelvisCatphan600 (CTP404) CBCT# to REDn/apCTDmean-PHANTOM<1.0% 3.0% (motion)n/sn/s
        CIRS002PRA (pelvic)
        Richter et al. (2008)
        • Richter A.
        • et al.
        Investigation of the usability of conebeam CT data sets for dose calculation.
        Elekta Synergy XVIADAC Pinnacle (v7.6s)n/sCatphan (CTP503)Thorax/PelvisCatphan (CTP503) CBCT# to PDn/apCTDmean-PHANTOM(1.0 ± 1.1; 3.2 ± 0.7)n/sn/s
        Head(0.8 ± 0.8; 4.2 ± 1.2)
        Petit et al. (2008)
        • Petit S.F.
        • Van Elmpt W.J.C.
        • Nijsten S.M.J.J.G.
        • Lambin P.
        • Dekker A.L.A.J.
        Calibration of megavoltage cone-beam CT for radiotherapy dose calculations: Correction of cupping artifacts and conversion of CT numbers to electron density.
        MVision Siemens MV CBCTXiO (v4.3.3)MGSSmall homogeneous cylinderLow-dose ratePhantom (water cilinder) CBCT# to EDn/apCTDmean-PHANTOM0.3n/sn/s
        Cupping0.2
        Large homogeneous cylindern/a7.1
        Cupping0.4
        Homogeneous IMRT phantomn/a1.0
        Cupping0.9
        Guan et al. (2009)
        • Guan H.
        • Dong H.
        Dose calculation accuracy using cone-beam CT (CBCT) for pelvic adaptive radiotherapy.
        Varian OBI (v1.4)EclipseAAAPelvic phantomPelvisNuclearAssociates76-430+CIRS002H9K CBCT# to REDn/apCTDmean-PTV(−2.0; 0.9)n/sDmean: Bld = (−1.4; 0.0), Rct = (−0.6;0.7)
        Eclipse Default CBCT# to RED(−0.7; −0.4)Dmean: Bld = −0.7, Rct = (−0.4;−0.3)
        Cathpan Vendor CBCT# to RED(−3.4; −2.4)Dmean: Bld = (−1.7; −1.2), Rct = (−0.9;−0.7)
        Catphan500+annulus CBCT# to RED(−4.9; −4.4)Dmean: Bld = (−2.0; -1.5), Rct = (−1.2;−0.9)
        Hatton et al. (2009)
        • Hatton J.
        • McCurdy B.
        • Greer P.B.
        Cone beam computerized tomography: The effect of calibration of the Hounsfield unit number to electron density on dose calculation accuracy for adaptive radiation therapy.
        Varian 2100 IX OBIEclipse (v8.6)AAASlabs phantomPelvisCatphan600 (CTP404) CBCT# to REDn/aIonization chamberDmean-SLABS<6.0n/sn/s
        CIRS062M CBCT# to RED<20.0
        CIRS062M+scattering CBCT# to RED<3.0
        RT01 trial semi-anthropomorphic phantomCatphan600 (CTP404) CBCT# to REDDmean-PTV<1.5
        CIRS062M CBCT# to RED<4.0
        CIRS062M + scattering CBCT# to RED<0.5
        Rong et al. (2010)
        • Rong Y.
        • Smilowitz J.
        • Tewatia D.
        • Tomé W.A.
        • Paliwal B.
        Dose Calculation on KV Cone Beam CT Images: An Investigation of the Hu-Density Conversion Stability and Dose Accuracy Using the Site-Specific Calibration.
        Varian TrilogyEclipse (v8.1)AAARANDO headNot defined (FF, 125kV, 80mA, 25ms)Inner cylinder CIRS062M + water blocksn/apCTDmean-PTV&OAR<1.0n/sDVH: LPar < 1.0, BS < 1.0
        Lung phantomNot defined (HF, 125kV, 80mA, 25ms)Inner cylinder CIRS062M + water blocks1.5n/s
        Outer ring CIRS062M + water blocks + low density material inner cylinder0.5
        CIRS062M + water blocks5.4
        Body phantomNot defined (HF, 125kV, 80mA, 25ms)Inner cylinder CIRS062M + water blocks2.1DVH: OARs < 2.0
        CIRS062M + water blocks0.0DVH: OARs = 1.0
        Petit et al. (2010)
        • Petit S.F.
        • van Elmpt W.J.C.
        • Lambin P.
        • Dekker A.L.A.J.
        Dose recalculation in megavoltage cone-beam CT for treatment evaluation: Removal of cupping and truncation artefacts in scans of the thorax and abdomen.
        Siemens MVisionXiO (v4.3.4)FSPCIRS thoraxLow doseCylinder filled with watern/apCTDmean-CENTRE−2.02%/2mm74.0n/s
        Cupping & truncation0.599.0
        RANDO thoraxn/a11.074.0
        Cupping & truncation0.099.0
        CIRS 002H5n/a6.070.0
        Cupping & truncation1.0100.0
        Wang et al. (2013)
        • Wang J.
        • et al.
        Using corrected cone-beam CT image for accelerated partial breast irradiation treatment dose verification: the preliminary experience.
        Elekta Synergy XVIPinnacle (v8.0)n/sAnthropomorphic thorax phantomNot defined (S20, 120kVp, 36.1mAs)CBCT# to PD
        • Hu W.
        • Ye J.
        • Wang J.
        • Ma X.
        • Zhang Z.
        Use of kilovoltage X-ray volume imaging in patient dose calculation for head-and-neck and partial brain radiation therapy.
        n/apCTDPTV-95%−0.8n/sDVH: IpsLng(V30) = 0.0, H(V5) = 0.0
        ScatterpCTDPTV-95%0.0n/sDVH: IpsLng(V30) = 0.0, H(V5) = 0.0, Br(V100) = −6.0
        Elstrøm et al. (2014)
        • Elstrom U.V.
        • Olsen S.R.K.K.
        • Muren L.P.
        • Petersen J.B.B.B.
        • Grau C.
        The impact of CBCT reconstruction and calibration for radiotherapy planning in the head and neck region-a phantom study.
        Varian Trilogy Tx OBI (v1.4)Eclipsen/sRANDO headStandard headPhantom (Gammex RMI467) CBCT# to REDn/apCTDmean-PTV−0.41%/3mm99.9Dmean: SC = −0.1, LPar = −0.1, RPar = −0.2, Mnd = −0.3
        Standard head FF−0.499.9Dmean: SC = −0.1, LPar = −0.2, RPar = −0.2, Mnd = −0.3
        High quality head−0.499.8Dmean: SC = −0.1, LPar = −0.2, RPar = −0.2, Mnd = −0.3
        High quality head FF−0.599.8Dmean: SC = −0.1, LPar = −0.2, RPar = −0.2, Mnd = −0.3
        Standard headPhantom (PMMA) CBCT# to RED0.199.9Dmean: SC = 0.1, LPar = 0.1, RPar = 0.0, Mnd = −0.1
        Standard head FF0.099.9Dmean: SC = 0.1, LPar = 0.1, RPar = 0.0, Mnd = −0.2
        High quality head0.199.9Dmean: SC = 0.1, LPar = 0.1, RPar = 0.0, Mnd = −0.1
        High quality head FF−0.199.9Dmean: SC = 0.0, LPar = 0.1, RPar = 0.0, Mnd = −0.2
        Standard headPhantom (CIRS062M) CBCT# to RED0.199.9Dmean: SC = 0.1, LPar = 0.2, RPar = 0.0, Mnd = 0.1
        Standard head FF0.099.9Dmean: SC = 0.1, LPar = 0.1, RPar = 0.0, Mnd = 0.0
        High quality head0.199.9Dmean: SC = 0.2, LPar = 0.2, RPar = 0.0, Mnd = 0.1
        High quality head FF0.199.9Dmean: SC = 0.1, LPar = 0.2, RPar = 0.0, Mnd = 0.1
        Ma et al. (2014)
        • Ma C.
        • Cao J.
        • Yin Y.
        • Zhu J.
        Radiotherapy dose calculation on KV cone-beam CT image for lung tumor using the CIRS calibration.
        Varian TrilogyPinnacle (v8.0)CCCIRS062MNot defined (HF, 120kV, 25mA, 40ms)CIRS062M CBCT# to REDn/apCTDmax4.0n/sn/s
        AuthorMake/modelTPSAlgorithmPhantomProtocolCBCT CalibrationArtefact CorrectionReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Barateau et al. (2015)
        • Barateau A.
        • et al.
        Dose calculation accuracy of different image value to density tables for cone-beam CT planning in head & neck and pelvic localizations.
        Varian OBI (v1.5)EclipseAAAATOM 701-GStandard HeadPhantom (Catphan 600 CTP404) CBCT# to REDn/apCT (CT# to RED Gammex RMI467)Dmean-PTV0.4n/sDmean: LPar = 0.21, RPar = −0.36, D2%: SC = 1.53
        Phantom (CIRS 062M head) CBCT# to RED−0.6Dmean: LPar = −0.41, RPar = −1.09, D2%: SC = −0.41
        Standard PelvisPhantom (CIRS 062M body) CBCT# to RED−1.1D50%: Bld = −1.06, Rct = −0.09
        Phantom (Gammex RMI467) CBCT# to RED0.4D50%: Bld = 0.00, Rct = 0.36
        Held et al. (2015)
        • Held M.
        • Sneed P.K.
        • Fogh S.E.
        • Pouliot J.
        • Morina O.
        Feasibility of MV CBCT-based treatment planning for urgent radiation therapy: Dosimetric accuracy of MV CBCT-based dose calculations.
        Siemens Artiste & In-Line kViewPinnacle (v9.2)n/sPelvic water phantom with insertsn/sPhantom (CIRS062+water cylinder) CBCT# to PDn/apCTDPHANTOM<2.0n/sD surface region∼5.0
        Held et al. (2016)
        • Held M.
        • et al.
        Assessment of image quality and dose calculation accuracy on kV CBCT, MV CBCT, and MV CT images for urgent palliative radiotherapy treatments.
        Varian Truebeamn/sn/sWater cylinderHeadVarian Truebeam CBCT# to PDn/apCTDmean-PHANTOM−0.3 ± 2.43%/3mm (30% threshold)97.3n/s
        Water pelvic phantomPelvis0.3 ± 0.899.4
        Head phantomHead1.9 ± 4.394.4
        Thorax phantomThorax0.6 ± 5.099.6
        Pelvic phantomPelvis−0.1 ± 4.7100.0
        Elekta VersaHDWater cylinderHead and Neck S20Phantom (Elekta head – small – and thorax/pelvis – large) CBCT# to PD−0.6 ± 2.797.2
        Water pelvic phantomPelvis M20−0.0 ± 0.899.3
        Head phantomHead and Neck S20−1.1 ± 6.299.7
        Thorax phantomChest M20−3.5 ± 15.197.9
        Pelvic phantomPelvis M200.6 ± 2.699.8
        Siemens MV ArtisteWater cylinderrFOVPhantom (Artiste head, thorax and pelvis) CBCT# to PD−0.7 ± 6.596.7
        Water pelvic phantomeFOV0.2 ± 5.499.1
        Head phantomrFOV−3.7 ± 12.596.7
        Thorax phantomeFOV−0.3 ± 6.899.5
        Pelvic phantomeFOV−0.2 ± 3.899.9
        Yohannes et al. (2016)
        • Yohannes I.
        • Prasetio H.
        • Kallis K.
        • Bert C.
        Dosimetric accuracy of the cone-beam CT-based treatment planning of the Vero system: a phantom study.
        Brainlab Vero SystemPinnacleCCIn-house cubeNot defined (80kV, 100mA)Phantom (Catphan 504 CTP404) CBCT# to PDn/aIonization chamberDmean-CHAMBER−4.83%/3mm54.3n/s
        Phantom (Gammex RMI467) CBCT# to PD−0.891.8
        Phantom (Tissue equivalent) CBCT# to PD−0.392.3
        Not defined (100kV, 100mA)Phantom (Catphan 504 CTP404) CBCT# to PD−4.757.3
        Phantom (Gammex RMI467) CBCT# to PD−0.893.0
        Phantom (Tissue equivalent) CBCT# to PD−0.694.5
        Not defined (120kV, 100mA)Phantom (Catphan 504 CTP404) CBCT# to PD−4.464.6
        Phantom (Gammex RMI467) CBCT# to PD−0.596.9
        Phantom (Tissue equivalent) CBCT# to PD−0.396.8
        Kaliyaperumal et al. (2017)

        Kaliyaperuma IV, et al., Study of Variation in Dose Calculation Accuracy Between kV Cone‑Beam Computed Tomography and kV fan‑Beam Computed Tomography, J Med Phys, 42(3), 2017, p. 171–180, doi: 10.4103/jmp.JMP.

        Varian Clinac IX OBIEclipse (v10.0)AAARW 3 platesn/sPhantom (Catphan CTP404) CBCT# to EDn/apCTDPHANTOM<1.2%n/sn/s
        Catphan (c-shape target)DPTV-98%−1.31%/1mm(96.0;100.0)Dmean: Normal Structures = −1.2
        Catphan (cylindrical target)2.4(96.0;100.0)Dmean: Normal Structures = 0.5
        Catphan (ring target)0.3(96.0;100.0)Dmean: Normal Structures = 0.2
        Catphan (build up region)0.8(96.0;100.0)Dmean: Normal Structures = −0.4
        Kaplan et al. (2018)
        • Kaplan L.P.
        • Elstrøm U.V.
        • Møller D.S.
        • Hoffmann L.
        Cone beam CT based dose calculation in the thorax region.
        Varian Clinac IX OBIEclipse (v13.7.14)AAAAlderson Radiation Therapy phantomLow dose thoraxPhantom (CIRS062M) CBCT# to EDn/apCTDPTV-98%−1.2n/sDmedian: Brnc = −2.7, H = −0.3, SC = 0.5, RLng = 0.15
        Jarema et al. (2019)
        • Jarema T.
        • Aland T.
        Using the iterative kV CBCT reconstruction on the Varian Halcyon linear accelerator for radiation therapy planning for pelvis patients.
        Varian HalcyonEclipse (v15.6.03)AAAPelvis (10)PelvisPhantom (CIRS062M) CBCT# to EDn/apCTDmean-PTV(−1.7; 3.0)1%/1mm (10% threshold)99.4 ± 0.2n/s
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; CC = collapsed cone; FSP = fast superposition XB; MGS = multi-grid superposition convolution; HF = half fan; FF = full fan; eFOV = extended field of view; rFOV = regular field of view; CT# = CT number; CBCT# = CBCT value; ED = electron density; RED = relative electron density; PD = physical density; pCT = planning CT; PTV = planning target volume; Dmax = maximum dose; Dmean = mean dose; Dmedian = median dose; DPTV-#% = dose delivered to #% of PTV; D#% =  minimum doses received by at least #% of the volume that received the largest dose; OAR = organ at risk; R = right; L = left; Bld = bladder; Rct = rectum; Par = parotid; SC = spinal cord; BS = brain stem; Mnd = mandible; Lng = lung; H = hearth; Br = breast; Brnc = bronchi; Ips = ipsilateral; n/s = not supplied; n/a = not applicable.
      • 3-
        HU override: The dose is calculated on the CBCT images after overriding the CBCT HU values with either HU density or CT numbers from the CT images. The patient and phantom studies retrieved from the literature adopting this method are listed in Table 3a, Table 3b, respectively.
        Table 3aDose calculation studies applying HU override on CBCT images for different patients’ tumour sites, in chronological order.
        AuthorMake/modelTPSAlgorithmSiteProtocolHU overrideArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Van Zijtveld et al. (2007)
        • van Zijtveld M.
        • Dirkx M.
        • Heijmen B.
        Correction of conebeam CT values using a planning CT for derivation of the ‘dose of the day’.
        Elekta Synergy XVIn/sPBH&N (2)n/sHUROIn/an/srCTn/s2%/2mm(93.0; 98.0)DVH
        Boggula et al. (2009)
        • Boggula R.
        • et al.
        A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
        Elekta Synergy XVICorvusPBProstate (3)PelvisAuto HUMLT (3 thresholds)n/aUser specified CT# to REDpCTn/s3%/3mm(86.3 ± 5.4;96.3 ± 4.5)n/s
        Hu et al. (2011)
        • Hu C.C.
        • et al.
        Practically acquired and modified cone-beam computed tomography images for accurate dose calculation in head and neck cancer.
        Elekta Synergy XVIPinnacle (v7.6c)n/sNPC (1)Not defined (120kVp, 32mA, 10ms)CBCT# to CT# (mathematical model)n/aUser specified CT# to PDpCTDmean-TV&OAR<3n/sDVH
        Fotina et al. (2012)
        • Fotina I.
        • Hopfgartner J.
        • Stock M.
        • Steininger T.
        • Lütgendorf-Caucig C.
        • Georg D.
        Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
        Elekta Synergy XVIiPlan (v4.5.0)XV MCProstate (10)Prostate (no bowtie)Auto HUMLT (7 thresholds)n/aUser specified CT# to EDPopulation-based CBCT# to EDDmedian-PTV1.9n/sDmean: Bld = 0.5, Rct = 0.1, FemHeads  = 0.0
        HUROI0.1Dmean: Bld = 0.3, Rct = 0.0, FemHeads  = 0.0
        Lung (10)Thorax (no bowtie)Auto HUMLT (7 thresholds)<5.0Dmax: SC = −0.2, OES = 0.1%, Dmean: Lng = 0.0
        HUROI<5.0Dmax: SC = 0.2, OES = −0.1%, Dmean: Lng = 0.1
        H&N (7)Head (no bowtie)Auto HUMLT (7 thresholds)<0.5Dmax: SC = −1.3, BS = 0.2, Dmean: Par = 0.4
        HUROI<−3.0Dmax: SC = −1.3, BS = 0.2, Dmean: Par = 0.4
        Poludniowski et al. (2012)
        • Poludniowski G.G.
        • Evans P.M.
        • Webb S.
        Cone beam computed tomography number errors and consequences for radiotherapy planning: An investigation of correction methods.
        Elekta Synergy XVIPinnacle (v9.0)n/sPelvis (6)PelvisPatient-specific CBCT# to CT#n/aUser specified CT# to PD
        • Poludniowski G.
        • Evans P.M.
        • Kavanagh A.
        • Webb S.
        Removal and effects of scatter-glare in cone-beam CT with an amorphous-silicon flat-panel detector.
        vCTDmean-PTV−2.8n/sDmean: Rct = −3.02, FemHeads  = −2.21
        Scatter (empirical)−1.0Dmean: Rct = −0.80, FemHeads  = −0.83
        Scatter
        • Poludniowski G.
        • Evans P.M.
        • Kavanagh A.
        • Webb S.
        Removal and effects of scatter-glare in cone-beam CT with an amorphous-silicon flat-panel detector.
        −0.2Dmean: Rct = −0.37, FemHeads  = 0.19
        Phantom CBCT# to CT#Scatter2.1Dmean: Rct = 1.84, FemHeads  = 2.33
        Brain (6)HeadPatient-specific CBCT# to CT#n/a−0.1Dmean: Ch = −0.30, OptNrv = −1.02
        Scatter (empirical)1.2Dmean: Ch = 1.05, OptNrv = 0.38
        Scatter
        • Poludniowski G.
        • Evans P.M.
        • Kavanagh A.
        • Webb S.
        Removal and effects of scatter-glare in cone-beam CT with an amorphous-silicon flat-panel detector.
        0.1Dmean: Ch = 0.42, OptNrv = 0.26
        Phantom CBCT# to CT#Scatter
        • Poludniowski G.
        • Evans P.M.
        • Kavanagh A.
        • Webb S.
        Removal and effects of scatter-glare in cone-beam CT with an amorphous-silicon flat-panel detector.
        −0.2Dmean: Ch = −0.18, OptNrv = 0.00
        Usui et al. (2013)
        • Usui K.
        • et al.
        Dose calculation with a cone beam CT image in image-guided radiation therapy.
        Varian Clinac 21EXEclipse (v8.9)AAALung (8)n/sPhantom (Gammex RMI467) CBCT# to CT#n/an/spCTDmean-TV3.9 ± 1.72%/2mm96.0 ± 2.8n/s
        Patient-specific CBCT# to CT#3.5  ± 3.894.5 ± 9.1
        HUMLT (3 thresholds)1.2 ±  0.891.1 ± 19.0
        Onozato et al. (2014)
        • Onozato Y.
        • et al.
        Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
        Varian OBI (v1.5)Eclipse (v8.6.15)AAAProstate (10)PelvisAuto HUMLT (3 thresholds)n/aUser specified CT# to EDpCTDmean-PTV0.6 ± 0.72%/2mm98.4Dmean: Bld = 1.10 ± 0.82, Rct = 0.79 ± 1.01
        HUHM0.5 ± 0.398.6Dmean: Bld = 0.86 ± 0.46, Rct = 0.31 ± 0.45
        Veiga et al. (2014)
        • Veiga C.
        • et al.
        Toward adaptive radiotherapy for head and neck patients: Feasibility study on using CT-to-CBCT deformable registration for ‘dose of the day’ calculations.
        Varian OBI (v1.4)EclipseAAAH&N (5)Low dose thoraxPhantom (Catphan 504) CBCT# to CT#n/an/srCTDmean(%pD)1.2 ± 0.22%/2mm90.3 ± 4.1DVH of one patient & histograms for BS, SC, LPar, RPar
        Dunlop et al. (2015)
        • Dunlop A.
        • et al.
        Comparison of CT number calibration techniques for CBCT-based dose calculation.
        Elekta Synergy XVI (v4.5)RayStation (v3.99)CCPelvis (4)Pelvis M10CBCT# to CT#
        • Poludniowski G.G.
        • Evans P.M.
        • Webb S.
        Cone beam computed tomography number errors and consequences for radiotherapy planning: An investigation of correction methods.
        n/aUser specified CT# to PD
        • Poludniowski G.
        • Evans P.M.
        • Kavanagh A.
        • Webb S.
        Removal and effects of scatter-glare in cone-beam CT with an amorphous-silicon flat-panel detector.
        pCTDmedian-CTV0.8n/sDmean: Bld = 0.4 – D50%: FemHeads  = 2.4, Rct = 1.4
        HU (water only)0.3Dmean: Bld = 0.1 – D50%: FemHeads  = 0.9, Rct = 0.2
        HUMLT (water & bone)−1.0Dmean: Bld = −1.0 – D50%: FemHeads  = −1.1, Rct = −1.0
        Auto HUMLT (6 thresholds)−1.3Dmean: Bld = −1.5, D50%: FemHeads  = −0.7, Rct = 0.7
        Lung (3)Chest S20CBCT# to CT#
        • Poludniowski G.G.
        • Evans P.M.
        • Webb S.
        Cone beam computed tomography number errors and consequences for radiotherapy planning: An investigation of correction methods.
        2.4Dmean: Lng = 2.9, H = −0.2, D2%: SC = 3.7
        HU (water only)−6.8Dmean: Lng = −7.8, H = −3.4, D2%: SC = −2.4
        HUMLT (water & lung)0.4Dmean: Lng = 0.4, H = −0.9, D2%: SC = 1.1
        Auto HUMLT (6 thresholds)−1.3Dmean: Lng = −1.8, H = −2.2, D2%: SC = −0.8
        H&N (4)Head S20CBCT# to CT#
        • Poludniowski G.G.
        • Evans P.M.
        • Webb S.
        Cone beam computed tomography number errors and consequences for radiotherapy planning: An investigation of correction methods.
        0.3Dmean: Par = 0.5, D2%: SC = 0.7, BS = 0.5
        HU (water only)0.4Dmean: Par = 0.4, D2%: SC = 1.4, BS = 0.5
        Auto HUMLT (6 thresholds)−0.1Dmean: Par = −0.3, D2%: SC = 0.5, BS = −0.5
        Veiga et al. (2015)
        • Veiga C.
        • et al.
        Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
        Varian OBI (v1.4)Eclipse (v10.8)n/sH&N (3)HeadPhantom (Catphan 504) CBCT# to CT#n/an/srCTDRMS(%pD) (photon plan)1.8 ± 1.6n/sDVH
        DRMS(%pD) (proton plan)(1.4 ± 0.5; 3.3 ± 0.5)
        Kurz et al. (2015)
        • Kurz C.
        • et al.
        Comparing cone-beam CT intensity correction methods for dose recalculation in adaptive intensity-modulated photon and proton therapy for head and neck cancer.
        Elekta Synergy XVI R4.5RayStation (v4.6)n/sH&N (9)n/sPopulation-based CBCT# to CT#n/an/srCTn/s3%/3mm (photon plan)(89.0; 97.0)DVH
        3%/3mm (proton plan)(80.0; 95.0)
        AuthorMake/modelTPSAlgorithmSiteProtocolHU overrideArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Almatani et al. (2016)
        • Almatani T.
        • Hugtenburg R.P.
        • Lewis R.D.
        • Barley S.E.
        • Edwards M.A.
        Automated algorithm for CBCT-based dose calculations of prostate radiotherapy with bilateral hip prostheses.
        Elekta Synergy XVI (v4.5)OMP (v4.3)PBProstate (1)n/sHUMLT (5 thresholds)n/aTPS CT# to EDpCTDmean-PTV−1.5n/sDmean: Bld = −48.6, Rct = 2.6
        CC−1.0Dmean: Bld = −48.8, Rct = −2.3
        MC−0.6Dmean: Bld = −47.1, Rct = −2.4
        PBAuto HUMLT (5 thresholds)−2.0Dmean: Bld = −49.0, Rct = −2.7
        CC−1.6Dmean: Bld = −49.2, Rct = −2.6
        MC−1.2Dmean: Bld = −47.4, Rct = −2.7
        Arai et al. (2017)
        • Arai K.
        • et al.
        Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
        Varian Clinac iX OBI 1.6XiO-MPBH&N (10)Not defined (FF, 100kV, 10mA)HUHMn/an/spCTDPTV-98%5.7 ± 4.13%/3mm94.0 ± 3.6Dmean: LPar = 3.56 ± 2.35
        Chen et al. (2017)
        • Chen S.
        • et al.
        Feasibility of CBCT-based dose with a patient-specific stepwise HU-to-density curve to determine time of replanning.
        Varian iX Trilogy / TruebeamRayStation (v5.0)n/sProstate (10)n/sPatient-specific auto HUMLT (6 thresholds)n/an/a (density overridden)pCTDmean-PTV0.2 ± 1.33%/3mm100.0 ± 0.0OAR 3%/3mm: (70–90% max dose) = 98.9 ± 2.5, ( < 70% max dose) = 95.8 ± 5.7
        Lung (10)−0.4 ± 1.196.1 ± 5.0OAR 3%/3mm: (70–90% max dose) = 98.7 ± 3.4, ( < 70% max dose) = 98.7 ± 2.2
        H&N (10)0.0 ± 0.698.3 ± 1.5OAR 3%/3mm: (70–90% max dose) = 92.9 ± 5.5, ( < 70% max dose) = 92.1 ± 7.2
        Pancreas (10)−0.2 ± 1.099.1 ± 2.4OAR 3%/3mm: (70–90% max dose) = 100.0 ± 0.0, ( < 70% max dose) = 96.9 ± 4.9
        MacFarlane et al. (2018)
        • Macfarlane M.
        • et al.
        Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment.
        Varian Clinac iX -Truebeam OBIPinnacle (v9.7)CCH&N (15)Standard & Low dose headHUROIInhomogeneityn/srCTDmean-Target−1.5 ± 0.83%/3mm94.4 ± 4.4Dmax(0.01cc): BS = −0.5 ± 1.5, SC = −3.4 ± 3.1, Dmean: LPar = 0.5 ± 1.9, RPar = −0.4 ± 2.4
        Giacometti et al. (2018)
        • Giacometti V.
        • et al.
        An evaluation of techniques for dose calculation on cone beam computed tomography.
        Varian TruebeamEclipse (v13.59)AAAProstate (5)PelvisAuto HUMLT (7 thresholds)n/an/a (density overridden)pCTDmedianPTV-99%0.32%/0.1mm (50% threshold)99.9Dmedian: Bld = 0.1, Rct = 0.1
        Lung (5)Thorax−0.694.9Dmedian: SC = 0.0, BS = 0.2
        H&N (5)Head0.195.5Dmax(0.01cc): SC = 0.0, Dmax(0.1cc): Oes = 0.0%, H = 0.1
        Schröder et al. (2019)

        Schröder L, Stankovic U, Remeijer P, Sonke J-J, Evaluating the impact of cone-beam computed tomography scatter mitigation strategies on radiotherapy dose calculation accuracy, Phys Imaging Radiat Oncol, 10, 2019, pp. 35–40, doi: 10.1016/j.phro.2019.04.001.

        Elekta Synergy XVI 5.0XiO (v4.3.4)n/sProstate (26)Not defined (M, 32mA, 40ms)Phantom (CIRS062M body) CBCT# to CT#Scatter (uniform)User specified CT# to PDvCTn/s2%/1mm (20% threshold)>95.0DVH
        Pelvis (28)>75.0
        Lung (24)Not defined (M,16mA, 40ms)Phantom (CIRS062M lung) CBCT# to CT#>50.0
        H&N (32)Not defined (S, 16mA, 40ms)Phantom (CIRS062M head) CBCT# to CT#>75.0
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; CC = collapsed cone; MC = Monte Carlo; H&N = head and neck; NPC =  nasopharyngeal carcinoma; HF = half fan; FF = full fan; CT# = CT number; CBCT# = CBCT value; MLT = multilevel-threshold; HM = histogram matching; ROI = region of interest; ED = electron density; RED = relative electron density; PD = physical density; pCT = planning CT; vCT = virtual CT; rCT = rescan CT; PTV = planning target volume; CTV = clinical target volume; Dmax = maximum dose; Dmean = mean dose; Dmedian = median dose; DPTV-#% = dose delivered to #% of PTV; Dmedian-PTV-#% = median dose delivered to #% of PTV; Dmean(%pD) = mean dose as percentage of the prescribed dose; DRMS(%pD) = root mean square as percentage of the prescribed dose; D#% =  minimum doses received by at least #% of the volume that received the largest dose; Dmax(#cc) = maximum dose to #cm3 of the PTV; OAR = organ at risk; R = right; L = left; Bld = bladder; Rct = rectum; Par = parotid; SC = spinal cord; Ch = chiasm; Oes = oesophagus; OptNrv = optic nerve; BS = brain stem; Lng = lung; H = hearth; Brnc = bronchi; Fem = femur; n/s = not supplied; n/a = not applicable.
        Table 3bDose calculation studies applying HU override on CBCT images for different phantoms, in chronological order.
        AuthorMake/modelTPSAlgorithmPhantomProtocolHU overrideArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Boggula et al. (2009)
        • Boggula R.
        • et al.
        A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
        Elekta Synergy XVICorvusPBPelvic phantomStandard PelvisAuto HUMLT (3 thresholds)n/aUser specified CT# to REDpCTn/s3%/3mm97.2 ± 0.7n/s
        Hu et al. (2011)
        • Hu C.C.
        • et al.
        Practically acquired and modified cone-beam computed tomography images for accurate dose calculation in head and neck cancer.
        Elekta Synergy XVIPinnacle (v7.6c)n/sCatphan 503Not defined (120kVp, 32mA, 10ms)Mathematical model for CBCT# to CT#n/aUser specified CT# to PDpCTDmean-PHANTOM(−0.2; 1.9)n/sn/s
        RANDO headDmean-PTV0.1Dmean: LPar = 1.5, RPar = 1.2, SC = −6.1, BS = −0.4%
        Usui et al. (2013)
        • Usui K.
        • et al.
        Dose calculation with a cone beam CT image in image-guided radiation therapy.
        Varian Clinac 21EXEclipse (v8.9)AAAI'mRT phantomn/sPhantom (Gammex RMI467) CBCT# to CT#n/an/spCTDmean-TARGET0.02%/2mm100.0n/s
        Phantom-specific CBCT# to CT#0.62%/2mm100.0
        HUMLT (3 thresholds)0.9100.0
        Onozato et al. (2014)
        • Onozato Y.
        • et al.
        Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
        Varian OBI (v1.5)Eclipse (v8.6.15)AAAPelvic phantomPelvisAuto HUMLT (3 thresholds)n/aUser specified CT# to EDpCTDmean-PTV0.2n/sn/sHistogram – Dmean: Bld = 0.1: Rct = 0.0
        HUHM0.1n/sn/sHistogram – Dmean: Bld = 0.2: Rct = 0.0
        Liu et al. (2015)

        Liu B, Lerma FA, Wu J, Yi BY, Yu C, Tissue Density Mapping of Cone Beam CT Images for Accurate Dose Calculations, Int J Med Phys, 2015, pp. 162–171, doi: 10.4236/ijmpcero.2015.42020.

        Varian TrilogyPinnacle (v8.1)n/sRANDO pelvisn/sHUMLT (5 thresholds)n/an/a (density overridden)pCTn/s2%/2mm(96.0; 98.0)n/s
        RANDO lung(97.0; 99.0)
        RANDO head(99.0; 100.0)
        Park et al. (2015)
        • Park Y.K.
        • Sharp G.C.
        • Phillips J.
        • Winey B.A.
        Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
        Elekta Synergy XVIXiOPBPelvic phantomNot defined (M15, 120kV, 40&64mA, 40ms)Phantom-specific (soft tissue-based) CBCT# to CT#n/aTPS CT# to RSPpCTn/s3%/3mm (10% threshold)94.8n/s
        Scatter (uniform)92.4
        Thorax phantomNot defined (M20, 120kV, 40&64mA, 40ms)n/a84.4
        Scatter (uniform)69.0
        Abdomen phantomNot defined (M20, 120kV, 40&64mA, 40ms)n/a96.4
        Scatter (uniform)97.5
        Head phantomNot defined (S20, 100kV, 10mA, 10&20ms)n/a85.6
        Scatter (uniform)87.7
        Arai et al. (2017)
        • Arai K.
        • et al.
        Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
        Varian Clinac iX OBI 1.6XiO-MPBPelvic phantomPelvisHUHMn/an/spCTDPTV-98%0.23%/3mm97.3Dmean: Rct = 0.14, Bld = 0.75
        H&N phantomNot defined (FF, 100kV, 10mA)2.799.7Dmean: LPar = 0.56
        Rafic et al. (2018)
        • Rafic K.M.
        • Amalan S.
        • Timothy Peace B.S.
        • Ravindran B.P.
        Extended localization and adaptive dose calculation using HU corrected cone beam CT: Phantom study.
        Varian OBI (v1.6)Eclipsen/sAlderson RANDONot defined (HF, 125kVp, 80mA)HU local correction – Stitch 1n/an/spCTDmean-TARGET(0.3; 0.7)3%/3mm (reference to global dose)(99.9; 100.0)Dmean: Bld = (1.2; 1.9), LFemHeads = (−1.5; −0.7), RFemHeads = (−0.8; 0.0), Vrt = (−0.2; 0.5)
        Octavius ion chambern/s(96.8; 99.2)n/s
        HU local correction – Stitch 2pCT(0.3; 0.6)(99.9; 100.0)Dmean: Bld = (1.7; 2.3), LFemHeads = (−2.9; −2.2), RFemHeads = 0.8; Vrt = (−2.0; −1.2)
        Octavius ion chambern/s(96.9; 99.2)n/s
        Schröder et al. (2019)

        Schröder L, Stankovic U, Remeijer P, Sonke J-J, Evaluating the impact of cone-beam computed tomography scatter mitigation strategies on radiotherapy dose calculation accuracy, Phys Imaging Radiat Oncol, 10, 2019, pp. 35–40, doi: 10.1016/j.phro.2019.04.001.

        Elekta Synergy XVI 5.0XiO (v4.3.4)n/sAlderson Rando (pelvis)Not defined (M, 32mA, 40ms)Phantom (CIRS062M body) CBCT# to CT#Scatter (uniform)User specified CT# to PDpCTDmean-PHANTOM3.4 ± 2.1n/sn/s
        Alderson Rando (prostate)2.4 ± 1.8
        Alderson Rando (lung)Not defined (S, 16mA, 40ms)Phantom (CIRS062M lung) CBCT# to CT#1.4 ± 2.8
        Alderson Rando (H&N)Not defined (M,16mA, 40ms)Phantom (CIRS062M head) CBCT# to CT#2.6 ± 10.1
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; H&N = head and neck; HF = half fan; FF = full fan; CT# = CT number; CBCT# = CBCT value; MLT = multilevel-threshold; HM = histogram matching; ED = electron density; RED = relative electron density; PD = physical density; RSP = relative stopping power; pCT = planning CT; PTV = planning target volume; CTV = clinical target volume; Dmax = maximum dose; Dmean = mean dose; DPTV-#% = dose delivered to #% of PTV; OAR = organ at risk; R = right; L = left; Bld = bladder; Rct = rectum; Par = parotid; SC = spinal cord; Vrt = vertebrae; BS = brain stem; Fem = femur; n/s = not supplied; n/a = not applicable.
      • 4-
        Deformable image registration (DIR): The dose is calculated on images generated by deforming the geometry of pCT/rCT images onto CBCT images via a deformation vector field (HU values are not modified). The patient and phantom studies retrieved from the literature adopting this method are listed in Table 4a, Table 4b, respectively.
        Table 4aDose calculation studies on different tumour sites in patients adopting deformable image registration (DIR), in chronological order.
        AuthorMake/modelTPSAlgorithmSiteProtocolDIR algorithmArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Yang et al. (2007)
        • Yang Y.
        • Schreibmann E.
        • Li T.
        • Wang C.
        • Xing L.
        Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation.
        Varian TrilogyEclipsePBProstate (3)n/sFree form BSplinen/aUser specified CT# to REDCBCTDmax-Prostate<2.0n/sn/s
        Lung (1)<5.0
        Disher et al. (2013)
        • Disher B.
        • Hajdok G.
        • Wang A.
        • Craig J.
        • Gaede S.
        • Battista J.J.
        Correction for ‘artificial’ electron disequilibrium due to cone-beam CT density errors: Implications for on-line adaptive stereotactic body radiation therapy of lung.
        Varian OBI (v1.4)PinnacleCCLung (3)ThoraxANIMALn/aUser specified CT# to REDpCTDPTV-95%(4.0; 13.0)n/sn/s
        Onozato et al. (2014)
        • Onozato Y.
        • et al.
        Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
        Varian OBI (v1.5)Eclipse (v8.6.15)AAAProstate (10)PelvisBspline (CBCT deformed topCT)n/aUser specified CT# to EDpCTDmean-PTV1.2 ± 0.52%/2mm98.1Dmean: Bld = 0.54 ± 0.42, Rct = 0.66 ± 0.58
        Veiga et al. (2014)
        • Veiga C.
        • et al.
        Toward adaptive radiotherapy for head and neck patients: Feasibility study on using CT-to-CBCT deformable registration for ‘dose of the day’ calculations.
        Varian OBI (v1.4)EclipseAAAH&N (5)Low dose thoraxBSpline (NiftyReg)n/an/srCTDPTV-95%0.7 ± 0.22%/2mm97.1 ± 4.3DVH of one patient & histograms for BS, SC, LPar, RPar
        Landry et al. (2015)
        • Landry G.
        • et al.
        Investigating CT to CBCT image registration for head and neck proton therapy as a tool for daily dose recalculation.
        Elekta Synergy XVI R4.5CERR-basedMCH&N (6)H&NMorphons (REGGUI, Matlab)n/aUser specified CT# to PDrCTn/s2%/2mm(83.0; 89.0)DVH: differences[Gy] rCT, physician and DIR contours
        Landry et al. (2015)
        • Landry G.
        • et al.
        Phantom based evaluation of CT to CBCT image registration for proton therapy dose recalculation.
        H&N (1)n/a3mm/2mm WET94.0n/s
        Kurz et al. (2015)
        • Kurz C.
        • et al.
        Comparing cone-beam CT intensity correction methods for dose recalculation in adaptive intensity-modulated photon and proton therapy for head and neck cancer.
        Elekta Synergy XVI R4.5RayStation (v4.6)n/sH&N (9)n/sMorphonsn/an/srCTn/s3%/3mm (photon plan)(95.0; 96.0)DVH of one patient
        3%/3mm (proton plan)(90.0; 95.0)
        Veiga et al. (2015)
        • Veiga C.
        • et al.
        Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
        Varian OBI (v1.4)Eclipse (v10.8)n/sH&N (3)HeadBSpline + Velocity field (NiftyReg)n/an/srCTDRMS(%pD) (photon plan)0.6 ± 0.1n/sDmean to OARs (SP, BS, Par) = 0.2 ± 0.1 & DVH of one patient
        DRMS(%pD) (proton plan)(1.0 ± 0.3; 2.6 ± 0.6)Dmean to OARs (SP, BS, Par) = (1.3 ± 2.5; 1.8 ± 1.7%) & DVH of one patient
        Kurz et al. (2016)
        • Kurz C.
        • et al.
        Investigating deformable image registration and scatter correction for CBCT-based dose calculation in adaptive IMPT.
        Elekta Synergy XVI or Versa HD (v4.5)RayStation (v4.6)n/sProstate (4)Not defined (M15, 120kVp, 64mA)Morphons (REGGUI, Matlab)Scatter (priori CT-based)
        • Park Y.K.
        • Sharp G.C.
        • Phillips J.
        • Winey B.A.
        Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
        ,
        • Niu T.
        • Sun M.
        • Star-Lack J.
        • Gao H.
        • Fan Q.
        • Zhu L.
        Shading correction for on-board cone-beam CT in radiation therapy using planning MDCT images.
        n/svCTn/s2%/2mm(99.9; 100.0)DVH
        H&N (4)Not defined (S20, 100kVp, 10mA)n/arCT(57.8; 94.4)n/s
        Scatter (priori CT-based)
        • Park Y.K.
        • Sharp G.C.
        • Phillips J.
        • Winey B.A.
        Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
        ,
        • Niu T.
        • Sun M.
        • Star-Lack J.
        • Gao H.
        • Fan Q.
        • Zhu L.
        Shading correction for on-board cone-beam CT in radiation therapy using planning MDCT images.
        rCT(58.1; 92.7)
        vCT(96.0; 98.6)
        Huang et al. (2017)
        • Huang P.
        • et al.
        Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone-beam CT.
        Varian OBIEclipse (v13.5)CCLiver RIDL (7)Not defined (HF, 120kVp, 40mA, 40ms)Gradient-based free formn/an/spCTDmean-NORMAL LIVER(−7.8;15.7)n/sn/s
        Liver RIDL (16)(−35.3; -12.1)
        Veiga et al. (2016, 2017)
        • Veiga C.
        • et al.
        A comprehensive evaluation of the accuracy of CBCT and deformable registration based dose calculation in lung proton therapy.
        IBA (v1.4)Eclipse (v11.0)n/sLung (20)Not defined (HF, 110kVp, 1142mAs)Morphonsn/aUser specified CT# to RSPrCTDmean(%Dmax)−0.5 ± 0.9n/sn/s
        Thing et al. (2017)
        • Thing R.S.
        • Bernchou U.
        • Hansen O.
        • Brink C.
        Accuracy of dose calculation based on artefact corrected Cone Beam CT images of lung cancer patients.
        Elekta Synergy XVI or Versa HD (v4.5-5.0)PinnacleCCLung (21)Clinical 4DBSpline (NiftyReg)n/aUser specified CT# to EDrCTn/s2%/2mm93.1Box Plot: median γ2%/2mm: Lng>80.0, H∼100.0, Oes∼100.0, SC∼100.0
        Comprehensive artefacts99.4Box Plots: median γ2%/2mm: Lng∼100.0, H∼100.0, Oes∼100.0, SC∼100.0
        Cole et al. (2018)
        • Cole A.J.
        • Veiga C.
        • Johnson U.
        • D’Souza D.
        • Lalli N.K.
        • McClelland J.R.
        Toward adaptive radiotherapy for lung patients: Feasibility study on deforming planning CT to CBCT to assess the impact of anatomical changes on dosimetry.
        Varian OBI (v1.4)EclipseAAALung (7)Low dose thoraxBspline + Velocity field (NiftyReg)n/an/srCTDmax-CTV+1cm0.7 ± 0.7n/sDmean: Lng-GTV(V20Gy) = 0.8 ± 1.2, H(V40Gy) = 2.0 ± 2.9, SCnl+5mm = 2.2 ± 3.3
        Giacometti et al. (2018)
        • Giacometti V.
        • et al.
        An evaluation of techniques for dose calculation on cone beam computed tomography.
        Varian TruebeamEclipse (v13.5)AAAProstate (5)Pelvisn/sn/aUser specified CT# to REDpCTDmedianPTV-99%0.12%/0.1mm (50% threshold)99.8Dmedian: Bld = 0.0, Rct = 0.1
        Lung (5)Thorax−0.395.7Dmedian: SC = 0.0, BS = −0.1
        H&N (5)Head0.096.0Dmax(0.01cc): SC = −0.2, Dmax(0.1cc): Oes = 0.1, H = −0.1
        MacFarlane et al. (2018)
        • Macfarlane M.
        • et al.
        Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment.
        Varian Clinac iX -Truebeam OBIPinnacle (v9.7)CCH&N (15)Standard & Low dose headFast-symetric Demon (Pinnacle)Inhomogeneityn/srCTDmean-TV−1.0 ± 0.83%/3mm96.1 ± 3.3Dmax(0.01cc): BS = 0.0 ± 1.2, SC = −3.0 ± 3.3, Dmean: LPar = 0.3 ± 2.4, RPar = 0.2 ± 3.1
        Marchant et al. (2018)

        Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

        Elekta Synergy XVI (v4.5-5.0)Pinnacle (v9.8)n/sPelvis (15)n/sBSpline (Elastix)n/an/spCTDmean-PTV0.1 ± 0.2n/sRct(V40Gy) = −0.16 ± 0.66
        Free form BSpline (NiftyReg)0.0 ± 0.2Rct(V40Gy) = −0.11 ± 0.52
        Lung (15)BSpline (Elastix)0.4 ± 0.8Dmax(1cc): SC = 0.31 ± 1.14
        Free form BSpline (NiftyReg)0.3 ± 0.7Dmax(1cc): SC = 0.15 ± 1.27
        H&N (14)BSpline (Elastix)0.2 ± 0.1Dmean: CPar = 0.28 ± 0.20
        Free form BSpline (NiftyReg)−0.1 ± 0.2Dmean: CPar = 0.00 ± 0.30
        Hansen et al. (2018)
        • Hansen D.C.
        • et al.
        ScatterNet: a convolutional neural network for cone-beam CT intensity correction.
        Elekta Synergy XVI (R5.0.2)RayStation (v4.99)CCProstate (8)Not defined (M20, 40mA)n/sShading (machine learning)n/sShading corrected CBCTn/s2% (50% threshold) (photon plan)100.0n/s
        PB2% (50% threshold) (photon plan)53.0
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; CC = collapsed cone; MC = Monte Carlo; H&N = head and neck; HF = half fan; CT# = CT number; ED = electron density; RED = relative electron density; PD = physical density; RSP = relative stopping power; pCT = planning CT; CBCT = cone beam CT; vCT = virtual CT; rCT = rescan CT; PTV = planning target volume; CTV = clinical target volume; GTV = gross tumour volume; Dmax = maximum dose; Dmean = mean dose; Dmedian = median dose; DPTV-#% = dose delivered to #% of PTV; Dmedian-PTV-#% = median dose delivered to #% of PTV; Dmean(%pD) = mean dose as percentage of the prescribed dose; DRMS(%pD) = root mean square as percentage of the prescribed dose; Dmax(#cc) = maximum dose to #cm3 of the PTV; Dmean(%Dmax) = mean dose normalized to the maximum dose; OAR = organ at risk; R = right; L = left; C = contra; Bld = bladder; Rct = rectum; Par = parotid; SC = spinal cord; SCnl = spinal canal; Oes = oesophagus; BS = brain stem; Lng = lung; SP = spine; H = hearth; H = hearth; V(#Gy) = volume receiving #Gy or more; n/s = not supplied; n/a = not applicable.
        Table 4bDose calculation studies on different phantoms adopting deformable image registration (DIR), in chronological order.
        AuthorMake/modelTPSAlgorithmPhantomProtocolDIRArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Landry et al. (2015)
        • Landry G.
        • et al.
        Phantom based evaluation of CT to CBCT image registration for proton therapy dose recalculation.
        Elekta Synergy XVI R4.5n/sMCPMMA deformable phantom modelling H&N anatomical changesn/sMorphons (REGGUI)n/aUser specified CT# to PDrCTAverage range<1.0n/sn/s
        Park et al. (2015)
        • Park Y.K.
        • Sharp G.C.
        • Phillips J.
        • Winey B.A.
        Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
        Elekta Synergy XVIXiOPBPelvic phantomNot defined (M15, 120kV, 40&64mA, 40ms)n/sScatter (priori CT-based)TPS CT# to RSPpCTn/s3%/3mm98.4n/s
        Thorax phantomNot defined (M20, 120kV, 40&64mA, 40ms)98.6
        Abdomen phantomNot defined (M20, 120kV, 40&64mA, 40ms)99.7
        Head phantomNot defined (S20, 100kV, 10mA, 10&20ms)99.5
        Niepel et al. (2018)
        • Niepel K.
        • et al.
        Feasibility of 4DCBCT-based proton dose calculation: An ex vivo porcine lung phantom study.
        Elekta Synergy XVI R4.5RayStation (v4.99)n/sMoving phantom based on ex vivo porcine lungsChestMorphons (REGGUI)n/aUser specified CT# to RSP4DCTn/s3%/3mm(89.0; 99.0)n/s
        TPS = treatment planning system; PB = pencil beam; MC = Monte Carlo; H&N = head and neck; CT# = CT number; ED = electron density; RED = relative electron density; PD = physical density; RSP = relative stopping power; pCT = planning CT; rCT = rescan CT; OAR = organ at risk.
      • 5-
        Dose deformation: The dose is calculated by deforming/mapping dose values from an irradiated geometry (CBCT) to a reference geometry (pCT), utilising a deformation vector field. The patient and phantom studies retrieved from the literature adopting this method are listed in Table 5a and 5b, respectively.
        Table 5Dose calculation studies on different tumour sites in patients via dose deformation by means of a deformation vector field, in chronological order.
        AuthorMake/modelTPSAlgorithmSite/PhantomProtocolDose WarpingArtefact CorrectionCalibration CurveReferenceDifferenceγ-analysis [%]Difference (OAR)
        Wen et al. (2012)
        • Wen N.
        • et al.
        Evaluation of the deformation and corresponding dosimetric implications in prostate cancer treatment.
        Varian TrilogyEclipsen/sProstate (5)n/sDIR + WARP (Bspline)n/an/spCTDmean-PTV−1.9 ± 0.1%n/sDmean[%]: Bld = (10.9; 42.4), Rct = (−3.4; 30.7)
        Garcia Molla et al. (2015)
        • García-Mollá R.
        • et al.
        Validation of a deformable image registration produced by a commercial treatment planning system in head and neck.
        Elekta Synergy XVI (v4.2.1)RayStation (v4.0.1.4)n/sH&N (5)n/sDIR + WARP (Hybrid algorithm)n/an/spCTDVH40 ± 40cGyn/sDVH[cGy]: TMJ = 99 ± 70, BS = 100 ± 100, B = 20 ± 20, Lrx = 50 ± 10, Jaw = 40 ± 30, SC = 110 ± 80, OptNrv = 36 ± 2, E = 20 ± 20, Par = 80 ± 40
        Veiga et al. (2015)
        • Veiga C.
        • et al.
        Toward adaptive radiotherapy for head and neck patients: Uncertainties in dose warping due to the choice of deformable registration algorithm.
        Varian OBI (v1.4)EclipseAAAH&N (5)Low dose thoraxDIR(sas) + WARP(sas)n/an/sDIR(svf) + WARP(svf)D99%(%pD)7.0 ± 3.0%n/sDmean[%] SC = 0.4 ± 0.6, BS = 0.6 ± 0.8 Par = 0.6 ± 0.3
        DIR(sas) + WARP(inv)7.0 ± 3.0%Dmean[%] SC = 0.4 ± 0.6, BS = 0.6 ± 0.8 Par = 0.3 ± 0.4
        DIR(ics) + WARP(ics)9.0 ± 5.0%Dmean[%] SC = 0.5 ± 0.4, BS = 0.5 ± 0.4 Par = 0.9 ± 0.5
        Veiga et al. (2016, 2017)
        • Veiga C.
        • et al.
        First Clinical Investigation of Cone Beam Computed Tomography and Deformable Registration for Adaptive Proton Therapy for Lung Cancer.
        ,
        • Veiga C.
        • et al.
        A comprehensive evaluation of the accuracy of CBCT and deformable registration based dose calculation in lung proton therapy.
        IBA (v1.4)Eclipse (v11.0)n/sLung (20)Not defined (HF, 110kVp, 1142mAs)DIR + WARP (Morphons)n/aUser specified CT# to RSPrCTDmean(%Dmax)−0.4 ± 1.0%n/sDVH of three patients
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; H&N = head and neck; HF = half fan; CT# = CT number; RSP = relative stopping power; pCT = planning CT; rCT = rescan CT; PTV = planning target volume; Dmean = mean dose; D#%(%pD)  =  minimum doses received by at least #% as percentage of the prescribed dose; Dmax(#cc) = maximum dose to #cm3 of the PTV; Dmean(%Dmax) = mean dose normalized to the maximum dose; OAR = organ at risk; R = right; L = left; Bld = bladder; Rct = rectum; B = brain; Par = parotid; SC = spinal cord; OptNrv = optic nerve; BS = brain stem; TMJ = temporomandibular joint; E =  eye; Lrx = larynx; Lng = lung; Sas = standard asymmetric registration; Inv = inverse; Ics = inverse-consistent symmetric; Svf = symmetric registration parameterized by a stationary velocity field; n/s = not supplied; n/a = not applicable.
      • 6-
        Combined techniques: The dose is calculated by adopting a combination of the listed techniques, including CBCT calibration, HU override and DIR. The patient and phantom studies retrieved from the literature adopting this method are listed in Table 6a, Table 6b, respectively.
        Table 6aDose calculation studies using a combination of techniques for different tumour sites in patients, in chronological order.
        AuthorMake/modelTPSAlgorithmSiteProtocolDose Calculation MethodArtefact CorrectionCalibration CurveReferenceDifference [%]γ-analysis [%]Difference (OAR) [%]
        Disher et al. (2013)
        • Disher B.
        • Hajdok G.
        • Wang A.
        • Craig J.
        • Gaede S.
        • Battista J.J.
        Correction for ‘artificial’ electron disequilibrium due to cone-beam CT density errors: Implications for on-line adaptive stereotactic body radiation therapy of lung.
        Varian OBI (v1.4)PinnacleCCLung (3)ThoraxDIR (CBCT to pCT) + CBCT calibration curven/aPhantom (Gammex RMI467) CBCT# to REDpCTDPTV-95%(−5.0; 6.0)n/sn/s
        DIR + Auto HUMLT (4 thresholds)User specified CT# to RED(1.0; 9.0)
        DIR + HUROI3.7
        Jin et al. (2013)
        • Jin X.
        • et al.
        CBCT-based volumetric and dosimetric variation evaluation of volumetric modulated arc radiotherapy in the treatment of nasopharyngeal cancer patients.
        Elekta SynergyRayStation (v3.5)n/sNPC (10)Not defined (120kVp, 25mA, 40ms)DIR (CBCT to pCT) + CBCT calibration curven/aPhantom (Catphan CTP503) CBCT# to PDpCTDGTV-95%<0.9n/sDmean: Lpar < 10.3, Rpar < 11.0,SC < 5.4, BS < 2.5
        Onozato et al. (2014)
        • Onozato Y.
        • et al.
        Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
        Varian OBI (v1.5)Eclipse (v8.6.15)AAAProstate (10)PelvisAuto HUMLT (4 thresholds) + DIRn/aUser specified CT# to EDpCTDmean-PTV0.8 ± 0.42%/2mm99.0Dmean: Bld = 0.55 ± 0.33, Rct = 0.61 ± 0.49
        HUHM + DIR0.8 ± 0.299.1Dmean: Bld = 0.26 ± 0.37, Rct = 0.39 ± 0.42
        Liu et al. (2015)

        Liu B, Lerma FA, Wu J, Yi BY, Yu C, Tissue Density Mapping of Cone Beam CT Images for Accurate Dose Calculations, Int J Med Phys, 2015, pp. 162–171, doi: 10.4236/ijmpcero.2015.42020.

        Varian Trilogyn/sn/sProstate (1)n/sDIR + HUMLT (5 thresholds)n/an/a (density overridden)pCTDmean < 1.0n/sDVH
        Lung (1)
        H&N (1)
        Arai et al. (2017)
        • Arai K.
        • et al.
        Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
        Varian Clinac iX OBI 1.6XiO-MPBH&N (10)Not defined (FF, 100kV, 10mA)HUHM + DIRn/an/spCTDPTV-98%5.7 ± 4.13%/3mm94.1 ± 3.5Dmean: LPar = 3.5 ± 2.4
        MacFarlane et al. (2018)
        • Macfarlane M.
        • et al.
        Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment.
        Varian Clinac iX Truebeam OBIPinnacle (v9.7)CCH&N (15)Standard & Low dose headDIR + Patient-specific CBCT# to CT#Inhomogeneityn/srCTDmean-TV−0.5 ± 0.83%/3mm95.0 ± 3.0Dmax(0.01cc): BS = 0.6 ± 1.0, SC = −2.0 ± 2.5, Dmean: LPar = 0.7 ± 1.5, RPar = 0.5 ± 2.2
        Landry et al. (2019)
        • Landry G.
        • et al.
        Comparing Unet training with three different datasets to correct CBCT images for prostate radiotherapy dose calculations.
        Elekta Synergy XVI (R5.0.2)RayStation (v4.6)CCProstate (8)Not defined (M20, 20ms, 20mA)CBCT# to CT#
        • Park Y.K.
        • Sharp G.C.
        • Phillips J.
        • Winey B.A.
        Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
        + DIR
        Intensity (machine learning)n/sShading corrected CBCT
        • Hansen D.C.
        • et al.
        ScatterNet: a convolutional neural network for cone-beam CT intensity correction.
        n/s2%/2mm(85.0; 100)n/s
        Schröder et al. (2019)

        Schröder L, Stankovic U, Remeijer P, Sonke J-J, Evaluating the impact of cone-beam computed tomography scatter mitigation strategies on radiotherapy dose calculation accuracy, Phys Imaging Radiat Oncol, 10, 2019, pp. 35–40, doi: 10.1016/j.phro.2019.04.001.

        Elekta Synergy XVI 5.0XiO (v4.3.4)n/sPelvis (28)Not defined (M, 32mA, 40ms)Anti-scatter grid + Phantom (CIRS062M body) CBCT# to CT#Scatter (uniform)User specified CT# to PDvCTn/s2%/1mm (20% threshold)> 90.0DVH
        Scatter (iterative)> 90.0
        Prostate (26)Scatter (uniform)> 85.0
        Scatter (iterative)> 95.0
        Lung (24)Not defined (M,16mA, 40ms)Anti-scatter grid + Phantom (CIRS062M lung) CBCT# to CT#Scatter (uniform)> 60.0
        Scatter (iterative)> 80.0
        H&N (32)Not defined (S, 16mA, 40ms)Anti-scatter grid + Phantom (CIRS062M head) CBCT# to CT#Scatter (uniform)> 90.0
        Scatter (iterative)> 85.0
        TPS = treatment planning system; AAA = analytical anisotropic algorithm; PB = pencil beam; CC = collapsed cone; H&N = head and neck; NPC =  nasopharyngeal carcinoma; FF = full fan; DIR = deformable image registration; MLT = multilevel-threshold; HM = histogram matching; ROI = region of interest; CT# = CT number; CBCT# = CBCT value; ED = electron density; RED = relative electron density; PD = physical density; pCT = planning CT; CBCT = cone beam; CT; vCT = virtual CT; rCT = rescan CT; PTV = planning target volume; GTV = gross tumour volume; Dmax = maximum dose; Dmean = mean dose; DPTV-#% = dose delivered to #% of PTV; DGTV-#% = dose delivered to #% of GTV; OAR = organ at risk; R = right; L = left; C = contra; Bld = bladder; Rct = rectum; Par = parotid; SC = spinal cord; BS = brain stem; n/s = not supplied; n/a = not applicable.
        Table 6bDose calculation studies using a combination of techniques for different phantoms, in chronological order.
        AuthorMake/modelTPSAlgorithmPhantomProtocolDose Calculation MethodArtefact CorrectionReferenceCalibration CurveDifference [%]γ-analysis [%]Difference OAR [%]
        Moteabbed et al. (2014)
        • Moteabbed M.
        • Sharp G.C.
        • Wang Y.
        • Trofimov A.
        • Efstathiou J.A.
        • Lu H.M.
        Validation of a deformable image registration technique for cone beam CT-based dose verification.
        Elekta Synergy XVIRayStationn/sPelvis full bladdern/sDIR on original images (Bspline/Demons) + DIR (Bspline)n/apCTn/sDmean-PROSTATE(2.9; 3.9)2%/2mm(84.8; 89.2)Dmean: Bld = (2.3; 2.9), Rct = (0.5; 2.9)
        DIR on original images (Bspline/Demons) + Phantom-specific CBCT# to CT#0.5100Dmean: Bld 0 = −0.2, Rct = (0.2; 0.4)
        DIR on original images (Bspline/Demons) + Phantom-specific CBCT# to CT# + DIR (BSpline)0.2100.0Dmean: Bld  = −0.3, Rct = (0.0; 0.1)
        Pelvis half full bladderDIR on original images (Bspline/Demons) + DIR (Bspline)(3.0; 3.2)(82.8; 89.3)Dmean: Bld  = (2.2; 3.0), Rct = (0.6; 2.2)
        DIR on original images (Bspline/Demons) + Phantom-specific CBCT# to CT#0.5100Dmean: Bld  = (−0.7; 0.2), Rct = (0.2; 0.3)
        DIR on original images (Bspline/Demons) + Phantom-specific CBCT# to CT# + DIR (BSpline)0.2100.0Dmean: Bld  = (0.2; 0.4), Rct = (0.0; 0.2)
        Pelvis empty bladderDIR on original images (Bspline/Demons) + DIR (Bspline)(3.2; 3.6)(81.7; 87.7)Dmean: Bld  = (7.8; 9.1), Rct = (1.1; 2.6)
        DIR on original images (Bspline/Demons) + Phantom-specific CBCT# to CT#0.6100Dmean: Bld  = (−0.1; 0.6), Rct = (0.3; 0.4)
        DIR on original images (Bspline/Demons) + Phantom-specific CBCT# to CT# + DIR (BSpline)0.2100.0Dmean: Bld  = (0.0; 1.4), Rct = (0.0; 0.2)
        Schröder et al. (2019)

        Schröder L, Stankovic U, Remeijer P, Sonke J-J, Evaluating the impact of cone-beam computed tomography scatter mitigation strategies on radiotherapy dose calculation accuracy, Phys Imaging Radiat Oncol, 10, 2019, pp. 35–40, doi: 10.1016/j.phro.2019.04.001.

        Elekta Synergy XVI 5.0XiO (v4.3.4)n/sAlderson Rando (pelvis)Not defined (M, 32mA, 40ms)Anti-scatter grid + Phantom (CIRS062M body) CBCT# to CT#Scatter (uniform)pCTUser specified CT# to PDDmean-PHANTOM2.0 ± 1.4n/sn/s
        Scatter (iterative)1.2 ± 1.1
        Alderson Rando (prostate)Scatter (uniform)1.1 ± 1.0
        Scatter (iterative)1.2 ± 1.1
        Alderson Rando (lung)Not defined (S, 16mA, 40ms)Anti-scatter grid + Phantom (CIRS062M lung) CBCT# to CT#Scatter (uniform)1.2 ± 1.6
        Scatter (iterative)1.2 ± 1.4
        Alderson Rando (H&N)Not defined (M, 16mA, 40ms)Anti-scatter grid + Phantom (CIRS062M head) CBCT# to CT#Scatter (uniform)1.0 ± 3.5
        Scatter (iterative)1.1 ± 3.6
        TPS = treatment planning system; H&N = head and neck; DIR = deformable image registration; PD = physical density; CT# = CT number; CBCT# = CBCT value; pCT = planning CT; CBCT = cone beam; CT; Dmean = mean dose; OAR = organ at risk; R = right; L = left; C = contra; Bld = bladder; Rct = rectum; n/s = not supplied; n/a = not applicable.
      Artefact correction can be applied on the CBCT images before adopting any of these dose calculation methods.
      Fig. 1 shows the number of patients (and matching publications) for whom the listed techniques were applied, grouped according to the tumour site. The main results and characteristics (CBCT device and protocol used, TPS and dose calculation algorithm adopted, patient/phantom imaged) reported in these studies are summarised in Tables 1-7.
      Figure thumbnail gr1
      Fig. 1Total number of patients for which the CBCT dose calculation was evaluated using different techniques (1 – pCT calibration, 2 – CBCT calibration, 3 – HU override, 4 – DIR, 5 – Dose deformation, 6 – Combined techniques). The number of publications matching the number of patients in each bar is reported in the table underneath the plot.
      Table 7Dose calculation studies for different tumour sites in patients and for different phantoms in tomotherapy MVCT adopting different methods, in chronological order.
      AuthorMake/modelTPSAlgorithmSite/PhantomProtocolMethodArtefact CorrectionReferenceCalibration CurveDifference [%]γ-analysis [%]Difference (OAR) [%]
      Langen et al. (2005)
      • Langen K.M.
      • et al.
      The use of megavoltage CT (MVCT) images for dose recomputations.
      TomoTherapy Hi·ArtTomotherapy Planning StationMCCIRS Thorax Phantom (rigid & distorted)n/sMVCT calibrationn/apCTPhantom (Gammex RMI467) MVCT# to REDDTarget-95%0.50n/sDVH
      Crop et al. (2012)
      • Crop F.
      • Bernard A.
      • Reynaert N.
      Improving dose calculations on tomotherapy MVCT images.
      TomoTherapy Hi·Art (v4.0.3)n/sn/sH&N Gammex RMI phantomn/sMVCT calibrationn/akV CTAuthor's model MVCT# to PDDPTV-50%0.20n/sn/s
      Pelvis (6)Phantom (Gammex RMI467) MVCT# to PDDDVH-50% < 3.20
      Author's model MVCT# to PD0.40
      H&N (6)Phantom (Gammex RMI467) MVCT# to PD < 1.40
      Author's model MVCT# to PD0.30
      Held et al. (2016)
      • Held M.
      • et al.
      Assessment of image quality and dose calculation accuracy on kV CBCT, MV CBCT, and MV CT images for urgent palliative radiotherapy treatments.
      Accuray TomoTherapyn/sn/sWater cylinder phantomn/sMVCT calibrationn/apCTUser specified MVCT# to PDDmean−0.4 ± 3.73%/3mm97.0n/s
      Water pelvis phantom−0.3 ± 0.6100.0
      Anthropomorphic head phantom−1.7 ± 6.199.5
      Anthropomorphic Thorax phantom−1.0 ± 9.099.2
      Anthropomorphic pelvic phantom0.2 ± 4.9100.0
      Branchini et al. (2017)
      • Branchini M.
      • et al.
      Validation of a method for ‘dose of the day’ calculation in head-neck tomotherapy by using planning ct-to-MVCT deformable image registration.
      Accuray TomoTherapyTomoHD System, DQA Station (v5.1.0.4)n/sH&N (10)n/sMVCT calibrationn/akV CTUser specified MVCT# to PDn/s2%/2mm95.7 ± 0.4n/s
      DIR (3B-SplinesFree Form)94.8 ± 0.8
      TPS = treatment planning system; MC = Monte Carlo; H&N = head and neck; DIR = deformable image registration; MVCT# = MVCT number; RED = relative electron density; PD = physical density; pCT = planning CT; PTV = planning target volume; Dmean = mean dose; OAR = organ at risk; n/s = not supplied; n/a = not applicable.

      3.1.1 pCT calibration

      Table 1a and Table 1b summarise the publications retrieved from the literature for patient and phantom studies in which dose was calculated directly on the CBCT images applying the calibration curve adopted for the pCT images used for planning. This methodology does not require any extra measurements because it calculates the attenuation corresponding to the CBCT HU values by utilising the calibration curve generated with CT numbers, generated with a different imaging modality (different energies, FOVs, collimation, etc). Although discrepancies between CBCT HU values and CT numbers are known, this method has been studied in 20 out of 69 cases investigating CBCT dose calculation.
      Elekta CBCT showed large differences between CBCT calculated and reference dose in a number of publications using this methodology [
      • Richter A.
      • et al.
      Investigation of the usability of conebeam CT data sets for dose calculation.
      ,
      • Boggula R.
      • et al.
      A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
      ,
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ,

      Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

      ,
      • Hu C.C.
      • et al.
      Practically acquired and modified cone-beam computed tomography images for accurate dose calculation in head and neck cancer.
      ] although dosimetric improvements up to 7.4% were observed when artefact corrections were applied [

      Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

      ,
      • Wang J.
      • et al.
      Using corrected cone-beam CT image for accelerated partial breast irradiation treatment dose verification: the preliminary experience.
      ].
      In almost all cases, differences from the reference dose using Varian CBCT have been within 1.5% [
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ,
      • Ding G.X.
      • et al.
      A study on adaptive IMRT treatment planning using kV cone-beam CT.
      ,
      • Onozato Y.
      • et al.
      Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Wang T.
      • et al.
      Dosimetric study on learning-based cone-beam CT correction in adaptive radiation therapy.
      ]. As expected, the largest differences were observed for lung images compared to the other tumour sites. This is most likely due to heterogeneities although this raises the question about the accuracy of different dose calculation algorithms. Another explanation is described in [
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ] where Elekta and Varian CT vs CBCT calibration curve are plotted. The two curves are almost overlapping for Varian while they are different for Elekta, thus explaining the higher dosimetric discrepancies.
      In the case of proton treatments, differences up to 27.1% in the dose to 98% of the PTV were observed for proton dose calculated on the acquired CBCT [
      • Arai K.
      • et al.
      Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
      ]. Authors attributed these differences to discrepancies between HU values in CBCT and pCT image that could cause a significant impact on the calculation of the proton range (and therefore in the dose calculation) when converted to relative proton stopping power.
      The advantage of using the pCT calibration approach is that it is fast and does not require any extra measurements. The disadvantage is that although this method may work for small objects, it is not reliable for larger and heterogeneous objects (and particularly for patient scans) due to the different amounts of scatter radiation present during the pCT and the CBCT acquisitions.

      3.1.2 CBCT calibration

      A new calibration curve with PD/RED/ED/SPR as a function of the HU values in the CBCT can be generated based on the CBCT images using phantom/patient/population-specific measurements (Table 2a, Table 2b). CT and CBCT calibration curves have been plotted and compared in several studies [
      • Richter A.
      • et al.
      Investigation of the usability of conebeam CT data sets for dose calculation.
      ,
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Barateau A.
      • et al.
      Dose calculation accuracy of different image value to density tables for cone-beam CT planning in head & neck and pelvic localizations.
      ,
      • Jin X.
      • et al.
      CBCT-based volumetric and dosimetric variation evaluation of volumetric modulated arc radiotherapy in the treatment of nasopharyngeal cancer patients.
      ,

      Schröder L, Stankovic U, Remeijer P, Sonke J-J, Evaluating the impact of cone-beam computed tomography scatter mitigation strategies on radiotherapy dose calculation accuracy, Phys Imaging Radiat Oncol, 10, 2019, pp. 35–40, doi: 10.1016/j.phro.2019.04.001.

      ]. While for negative HU values (lower density tissue) CT and CBCT calibration curves are in good agreement, for positive HU values the calibration curves are quite different, independent of the treated site.
      Several phantoms, briefly described in the supplementary material (Table S3), have been used to generate CBCT calibration curves. Errors up to 20.0% were calculated when using diagnostic CT scanners quality assurance phantoms [
      • Hatton J.
      • McCurdy B.
      • Greer P.B.
      Cone beam computerized tomography: The effect of calibration of the Hounsfield unit number to electron density on dose calculation accuracy for adaptive radiation therapy.
      ,
      • Richter A.
      • et al.
      Investigation of the usability of conebeam CT data sets for dose calculation.
      ,
      • Guan H.
      • Dong H.
      Dose calculation accuracy using cone-beam CT (CBCT) for pelvic adaptive radiotherapy.
      ,
      • Kaplan L.P.
      • Elstrøm U.V.
      • Møller D.S.
      • Hoffmann L.
      Cone beam CT based dose calculation in the thorax region.
      ,

      Kaliyaperuma IV, et al., Study of Variation in Dose Calculation Accuracy Between kV Cone‑Beam Computed Tomography and kV fan‑Beam Computed Tomography, J Med Phys, 42(3), 2017, p. 171–180, doi: 10.4103/jmp.JMP.

      ,
      • Jarema T.
      • Aland T.
      Using the iterative kV CBCT reconstruction on the Varian Halcyon linear accelerator for radiation therapy planning for pelvis patients.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ]. This can be attribute to the size of these phantoms (about 15 cm in diameter), which, due to the collimation used in standard CT scans, are of limited depth. If used for CBCT measurements, the imaging field can extend beyond the depth of the phantom and thus underestimate the scatter contribution which occurs during a patient CBCT acquisition. This lack of scatter causes dosimetric inaccuracies when undertaking dose calculations for objects/patients larger than the phantom size (diameter and depth).
      Commercial phantoms are now available that provide additional material perpendicular to the CT slices (depth), although a number of authors [
      • Hatton J.
      • McCurdy B.
      • Greer P.B.
      Cone beam computerized tomography: The effect of calibration of the Hounsfield unit number to electron density on dose calculation accuracy for adaptive radiation therapy.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Rong Y.
      • Smilowitz J.
      • Tewatia D.
      • Tomé W.A.
      • Paliwal B.
      Dose Calculation on KV Cone Beam CT Images: An Investigation of the Hu-Density Conversion Stability and Dose Accuracy Using the Site-Specific Calibration.
      ,
      • Ma C.
      • Cao J.
      • Yin Y.
      • Zhu J.
      Radiotherapy dose calculation on KV cone-beam CT image for lung tumor using the CIRS calibration.
      ] have reported mimicked this by adding additional material either side of the standard electron density phantoms described above. Differences from the reference dose of less than 0.8% were reported using the Gammex RMI467 phantom [
      • Yohannes I.
      • Prasetio H.
      • Kallis K.
      • Bert C.
      Dosimetric accuracy of the cone-beam CT-based treatment planning of the Vero system: a phantom study.
      ,
      • Barateau A.
      • et al.
      Dose calculation accuracy of different image value to density tables for cone-beam CT planning in head & neck and pelvic localizations.
      ,
      • Elstrom U.V.
      • Olsen S.R.K.K.
      • Muren L.P.
      • Petersen J.B.B.B.
      • Grau C.
      The impact of CBCT reconstruction and calibration for radiotherapy planning in the head and neck region-a phantom study.
      ] and generally less than 2.0% using CIRS062M phantom when used with extra slabs of water equivalent material to increase the scatter [
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Rong Y.
      • Smilowitz J.
      • Tewatia D.
      • Tomé W.A.
      • Paliwal B.
      Dose Calculation on KV Cone Beam CT Images: An Investigation of the Hu-Density Conversion Stability and Dose Accuracy Using the Site-Specific Calibration.
      ]. Hatton et al. [
      • Hatton J.
      • McCurdy B.
      • Greer P.B.
      Cone beam computerized tomography: The effect of calibration of the Hounsfield unit number to electron density on dose calculation accuracy for adaptive radiation therapy.
      ] reported mean dose differences up to 22.0% from ionisation chamber measurements when applying a calibration curve from a large diameter phantom to a smaller object, and up to 12.0% when a calibration curve from a small phantom was used for a larger object. This suggests that a site-specific approach may be more appropriate than a one-size-fits-all solution. Rong et al. [
      • Rong Y.
      • Smilowitz J.
      • Tewatia D.
      • Tomé W.A.
      • Paliwal B.
      Dose Calculation on KV Cone Beam CT Images: An Investigation of the Hu-Density Conversion Stability and Dose Accuracy Using the Site-Specific Calibration.
      ] confirmed that the HU uniformity deterioration increases with the size of the scanned objects, mostly due to scatter contamination, and the authors claimed this issue could be addressed using smaller cone angles. Indeed, differences from the reference dose between −1.2% and 1.0% were calculated when using site-specific CBCT phantoms for the calibration [
      • Yohannes I.
      • Prasetio H.
      • Kallis K.
      • Bert C.
      Dosimetric accuracy of the cone-beam CT-based treatment planning of the Vero system: a phantom study.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Rong Y.
      • Smilowitz J.
      • Tewatia D.
      • Tomé W.A.
      • Paliwal B.
      Dose Calculation on KV Cone Beam CT Images: An Investigation of the Hu-Density Conversion Stability and Dose Accuracy Using the Site-Specific Calibration.
      ], better mimicking the radiation response of the corresponding real tissues in both phantom studies and patient studies.
      According to Richter et al. [
      • Richter A.
      • et al.
      Investigation of the usability of conebeam CT data sets for dose calculation.
      ], the CBCT dose calculation accuracy could be further improved using patient-specific or population-based CBCT calibration curves rather than phantom ones, and this was later confirmed in [
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ]. The differences from the reference dose were generally less than 2.0% and there were minimal differences between patient-specific and population-based dose calculations.
      No trend can be observed between different CBCT models, although Siemens MVision reported the highest difference from the reference dose: up to 11.0% for the uncorrected CBCT that noticeably improved (up to 2.0%) with artefact correction [
      • Petit S.F.
      • van Elmpt W.J.C.
      • Lambin P.
      • Dekker A.L.A.J.
      Dose recalculation in megavoltage cone-beam CT for treatment evaluation: Removal of cupping and truncation artefacts in scans of the thorax and abdomen.
      ].
      No proton studies were found in the literature which reported using CBCT calibration curves for dose calculations.
      The CBCT calibration approach shows better results than the pCT calibration method, particularly for patient studies. Patient-specific CBCT calibration curves provided better results than using calibration phantoms [
      • Richter A.
      • et al.
      Investigation of the usability of conebeam CT data sets for dose calculation.
      ,
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ,
      • Hu W.
      • Ye J.
      • Wang J.
      • Ma X.
      • Zhang Z.
      Use of kilovoltage X-ray volume imaging in patient dose calculation for head-and-neck and partial brain radiation therapy.
      ], but the disadvantage of this approach is not generalizable and is carried out on a patient by patient basis. It is therefore time consuming and not clinically feasible without automation.

      3.1.3 HU override

      Another aspect to take into account when performing dose calculation based on CBCT is the variability of the HU values in the acquired images, which is attributable to acquisition parameters, object size and presence of inhomogeneities in the object itself. In addition, the increased scatter radiation can also cause an incorrect calculation of the photon attenuation, and consequently of the correspondent HU value and density. Thomas et al. [
      • Thomas S.J.
      Relative electron density calibration of CT scanners for radiotherapy treatment planning.
      ] reported that a difference of 8.0% in bone electron density corresponds to a dosimetric error of 1.0% for typical radiotherapy beams. A good indicator of the performance of a CBCT system is the stability of HU over time. Schröeder et al. [

      Schröder L, Stankovic U, Sonke JJ, Technical Note: Long-term stability of Hounsfield unit calibration for cone beam computed tomography, Med Phys, 0(25 cm), 2020, pp. 1–5, doi: 10.1002/mp.14015.

      ] investigated the HU reproducibility for an Elekta CBCT scanner (Synergy, XVI 5.0) over a period of 10 months, finding that the dosimetric change caused by the HU variability was less than 1.0%.
      To reduce the influence of CBCT artefacts and HU variability on the dose calculation, several studies have suggested overriding the density of the structures in the CBCT image with average HU values from the corresponding structures on the pCT/rCT [
      • Boggula R.
      • et al.
      A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
      ,
      • Onozato Y.
      • et al.
      Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
      ,
      • Rafic K.M.
      • Amalan S.
      • Timothy Peace B.S.
      • Ravindran B.P.
      Extended localization and adaptive dose calculation using HU corrected cone beam CT: Phantom study.
      ,
      • Macfarlane M.
      • et al.
      Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Arai K.
      • et al.
      Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ,
      • van Zijtveld M.
      • Dirkx M.
      • Heijmen B.
      Correction of conebeam CT values using a planning CT for derivation of the ‘dose of the day’.
      ,
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ,
      • Almatani T.
      • Hugtenburg R.P.
      • Lewis R.D.
      • Barley S.E.
      • Edwards M.A.
      Automated algorithm for CBCT-based dose calculations of prostate radiotherapy with bilateral hip prostheses.
      ,
      • Chen S.
      • et al.
      Feasibility of CBCT-based dose with a patient-specific stepwise HU-to-density curve to determine time of replanning.
      ,

      Liu B, Lerma FA, Wu J, Yi BY, Yu C, Tissue Density Mapping of Cone Beam CT Images for Accurate Dose Calculations, Int J Med Phys, 2015, pp. 162–171, doi: 10.4236/ijmpcero.2015.42020.

      ]. Some other studies generated a linear conversion curve to match the HU values in the CBCT images with the HU values in the pCT images [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ,
      • Hu C.C.
      • et al.
      Practically acquired and modified cone-beam computed tomography images for accurate dose calculation in head and neck cancer.
      ,
      • Park Y.K.
      • Sharp G.C.
      • Phillips J.
      • Winey B.A.
      Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
      ,
      • Usui K.
      • et al.
      Dose calculation with a cone beam CT image in image-guided radiation therapy.
      ,
      • Poludniowski G.G.
      • Evans P.M.
      • Webb S.
      Cone beam computed tomography number errors and consequences for radiotherapy planning: An investigation of correction methods.
      ,
      • Veiga C.
      • et al.
      Toward adaptive radiotherapy for head and neck patients: Feasibility study on using CT-to-CBCT deformable registration for ‘dose of the day’ calculations.
      ,
      • Kurz C.
      • et al.
      Comparing cone-beam CT intensity correction methods for dose recalculation in adaptive intensity-modulated photon and proton therapy for head and neck cancer.
      ] (Table 3a, Table 3b). In many publications, the CBCT HU values to CT numbers conversion curves are labelled “calibration curve”, not to be confused here with the CBCT calibration curve mentioned in the Section 3.1.2.
      Simple methods, such as applying a single HU value, equivalent to water, to the entire outline have been shown to give small median dose differences to the CTV for pelvis (0.3%) and head and neck (0.4%) but larger differences of 6.8% in lung patients [
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ]. Using an additional “lung” HU override combined with the water override, reduced this difference to 0.4%.
      Thresholding tools are available in the TPS to allow automatic contouring of structures with different densities. The segmentation of the images can be done using multilevel-threshold (MLT), based on algorithms that classify CBCT voxels with similar HUs into several ranges, based on the grey level. The thresholds can be generated automatically using specific software (here referred as “Auto HUMLT”) [
      • Boggula R.
      • et al.
      A new strategy for online adaptive prostate radiotherapy based on cone-beam CT.
      ,
      • Onozato Y.
      • et al.
      Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ,
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ,
      • Almatani T.
      • Hugtenburg R.P.
      • Lewis R.D.
      • Barley S.E.
      • Edwards M.A.
      Automated algorithm for CBCT-based dose calculations of prostate radiotherapy with bilateral hip prostheses.
      ,
      • Chen S.
      • et al.
      Feasibility of CBCT-based dose with a patient-specific stepwise HU-to-density curve to determine time of replanning.
      ] or manually [
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ,
      • Almatani T.
      • Hugtenburg R.P.
      • Lewis R.D.
      • Barley S.E.
      • Edwards M.A.
      Automated algorithm for CBCT-based dose calculations of prostate radiotherapy with bilateral hip prostheses.
      ,

      Liu B, Lerma FA, Wu J, Yi BY, Yu C, Tissue Density Mapping of Cone Beam CT Images for Accurate Dose Calculations, Int J Med Phys, 2015, pp. 162–171, doi: 10.4236/ijmpcero.2015.42020.

      ,
      • Usui K.
      • et al.
      Dose calculation with a cone beam CT image in image-guided radiation therapy.
      ]. Dunlop et al. [
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ] recommended the use Auto HUMLT rather than segmenting manually individual tissue (e.g. water/bone or water only), because it is faster, simpler and therefore easier to integrate into a typical clinical workflow. Better results were reported for head and neck and prostate cancer, compared to thorax and abdomen [
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ,
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ,
      • Chen S.
      • et al.
      Feasibility of CBCT-based dose with a patient-specific stepwise HU-to-density curve to determine time of replanning.
      ].
      Alternatively, a direct match between CBCT and pCT image cumulative histograms (HUHM) was proposed by Onozato et al. [
      • Onozato Y.
      • et al.
      Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
      ]: pCT and CBCT HU values were normalised, so that the HU values of the CBCT images matched the HU values in the pCT histogram. In this study, HUMLT and HUHM yielded to similar results both for prostate patients and for a pelvic phantom.
      For proton treatments, the HUHM method was employed by Arai et al. [
      • Arai K.
      • et al.
      Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
      ] for a proton plan for head and neck patients revealed a dose difference of 5.7% to 98.0% of PTV when performing a CBCT dose calculation. According to the investigation conducted by the authors, in regions with large artefacts, HUHM could not improve the CBCT dosimetric accuracy, thus generating potential problems in making histograms of HU values in proton beam therapy. Improved results were observed when the dose calculation was performed on a CBCT for pelvic (0.2%) and head and neck (2.7%) phantoms [
      • Arai K.
      • et al.
      Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
      ].
      Another possible option, here referred as “HUROI”, consists in segmenting structures (ROI’s) in the pCT and then transferring them to the CBCT images. The HU in the delineated ROI’s is manually assigned to the mean HU value of the same regions in the pCT images [
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ,
      • van Zijtveld M.
      • Dirkx M.
      • Heijmen B.
      Correction of conebeam CT values using a planning CT for derivation of the ‘dose of the day’.
      ,
      • Macfarlane M.
      • et al.
      Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment.
      ]. The direct comparison between HUMLT and HUROI presented by Fotina et al. [
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ] shows that similar results were obtained for prostate and lung, but for head and neck the results are 2.5% more accurate with HUMLT.
      For proton treatments, independently of whether HUMLD, HUHM, or HUROI are adopted, when applying the HU override technique in proton studies, similar gamma pass rates to photon studies were reported [
      • Park Y.K.
      • Sharp G.C.
      • Phillips J.
      • Winey B.A.
      Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.
      ,
      • Kurz C.
      • et al.
      Comparing cone-beam CT intensity correction methods for dose recalculation in adaptive intensity-modulated photon and proton therapy for head and neck cancer.
      ] but slightly higher dose differences to 98% of PTV (up to 5.7%) [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ] and root mean square of the dose difference (up to 3.3%) were measured. According to the authors, these differences can be explained by the limited FOV of CBCT, missing important geometrical information necessary for adaptive proton therapy and by the higher proton plans sensitivity to inaccuracies in the image registration. The poor results for the caudal head and neck cases in [
      • Kurz C.
      • et al.
      Comparing cone-beam CT intensity correction methods for dose recalculation in adaptive intensity-modulated photon and proton therapy for head and neck cancer.
      ] were explained by the artefacts and shadowing in the shoulder regions causing dosimetric inaccuracies both in IMRT (hotspots) and IMPT (distortions) plans.
      Stitching CBCT techniques to cover an entire phantom body were investigated by Rafic et al. [
      • Rafic K.M.
      • Amalan S.
      • Timothy Peace B.S.
      • Ravindran B.P.
      Extended localization and adaptive dose calculation using HU corrected cone beam CT: Phantom study.
      ] using an anthropomorphic phantom: mean dose differences<0.7% and gamma pass rates above 99.8% were measured. This is an interesting development as often a limitation of CBCT calculations are the limit to the FOV in the cranial-caudal direction.
      HU override techniques show results comparable to those achieved with the CBCT calibration method [
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ]. However, this method can provide unreliable dose distributions when a large variation in densities occurs within a segmented tissues (e.g. lungs) [
      • De Smet M.
      • Schuring D.
      • Nijsten S.
      • Verhaegen F.
      Accuracy of dose calculations on kV cone beam CT images of lung cancer patients.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Fotina I.
      • Hopfgartner J.
      • Stock M.
      • Steininger T.
      • Lütgendorf-Caucig C.
      • Georg D.
      Feasibility of CBCT-based dose calculation: Comparative analysis of HU adjustment techniques.
      ,
      • Dunlop A.
      • et al.
      Comparison of CT number calibration techniques for CBCT-based dose calculation.
      ].

      3.1.4 Deformable image registration

      Deformable image registration (DIR) is a technique that consists in geometrically deforming one image into another applying mathematical algorithms (eg. B-spline, Demons, etc). For CBCT dose calculation purposes, DIR is used to deform either the original pCT or the rescan rCT to match one of the patient’s CBCT images [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ,
      • Veiga C.
      • et al.
      A comprehensive evaluation of the accuracy of CBCT and deformable registration based dose calculation in lung proton therapy.
      ,

      Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

      ,
      • Onozato Y.
      • et al.
      Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Macfarlane M.
      • et al.
      Patient-specific calibration of cone-beam computed tomography data sets for radiotherapy dose calculations and treatment plan assessment.
      ,
      • Veiga C.
      • et al.
      Toward adaptive radiotherapy for head and neck patients: Feasibility study on using CT-to-CBCT deformable registration for ‘dose of the day’ calculations.
      ,
      • Thing R.S.
      • Bernchou U.
      • Hansen O.
      • Brink C.
      Accuracy of dose calculation based on artefact corrected Cone Beam CT images of lung cancer patients.
      ,
      • Cole A.J.
      • Veiga C.
      • Johnson U.
      • D’Souza D.
      • Lalli N.K.
      • McClelland J.R.
      Toward adaptive radiotherapy for lung patients: Feasibility study on deforming planning CT to CBCT to assess the impact of anatomical changes on dosimetry.
      ,
      • Landry G.
      • et al.
      Phantom based evaluation of CT to CBCT image registration for proton therapy dose recalculation.
      ,
      • Landry G.
      • et al.
      Investigating CT to CBCT image registration for head and neck proton therapy as a tool for daily dose recalculation.
      ,
      • Disher B.
      • Hajdok G.
      • Wang A.
      • Craig J.
      • Gaede S.
      • Battista J.J.
      Correction for ‘artificial’ electron disequilibrium due to cone-beam CT density errors: Implications for on-line adaptive stereotactic body radiation therapy of lung.
      ,
      • Kurz C.
      • et al.
      Investigating deformable image registration and scatter correction for CBCT-based dose calculation in adaptive IMPT.
      ,
      • Huang P.
      • et al.
      Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone-beam CT.
      ]. Thus, a deformation vector field (DVF) defining the voxel-to-voxel mapping between pCT and CBCT is generated. The outlined structures are propagated from pCT to CBCT according to the DVF and the new deformed image are referred to as vCT. Finally, the pCT plan is applied to the vCT and the dose delivered is calculated.
      DIR allows monitoring the organ deformation during the course of the treatment, particularly when daily or weekly CBCT scans are acquired. If anatomical changes, intra- or inter-fraction, are not corrected, they can lead to underdosage or overdosage to the target and OARs when using the planning dose distribution. Therefore, DIR can be used to propagate the planning segmentations to the images acquired on the day of the treatment (e.g. CBCT), thus giving an estimate of the dose delivered to the deformable structures and a prediction of the toxicity to the OARs [
      • Rigaud B.
      • et al.
      Deformable image registration for radiation therapy: principle, methods, applications and evaluation.
      ].
      For all the previously described methods, the correct way of calculating the dose delivered to the CBCT was to apply the pCT plan to the CBCT images. When performing DIR, a meaningful comparison between pCT and CBCT dose calculation can be reported only if small anatomical changes occurred in the patient between acquiring the pCT and acquiring the CBCT, otherwise it is not possible to establish which one is the actual reference image and dose. Thus, for most of the studies reported here using the DIR method (Table 4a), the reference dose is calculated on an rCT acquired on a date close to the CBCT. The rest of the studies chose pCT as reference because the CBCT images were chosen to match closely the pCT [
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ] or the contours were copied from the pCT as in [
      • Disher B.
      • Hajdok G.
      • Wang A.
      • Craig J.
      • Gaede S.
      • Battista J.J.
      Correction for ‘artificial’ electron disequilibrium due to cone-beam CT density errors: Implications for on-line adaptive stereotactic body radiation therapy of lung.
      ]. In other cases it was explicitly stated that a larger spread of errors were due to larger discrepancies between vCT and CBCT anatomy [

      Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

      ], or that pCT was chosen as a reference to iteratively compare the planned objectives while the treatment was progressing [
      • Huang P.
      • et al.
      Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone-beam CT.
      ].
      DIR tools are available in many TPS’s; one of DIR main advantages is that vCT images are not affected by HU inaccuracies because the original CT numbers are not modified by the deformation. On the other hand, one of DIR main challenges is its quantitative verification, often not possible because of the difficulty in determining the ground truth. In the Task Group 132 [
      • Brock K.K.
      • Mutic S.
      • McNutt T.R.
      • Li H.
      • Kessler M.L.
      Use of image registration and fusion algorithms and techniques in radiotherapy: Report of the AAPM Radiation Therapy Committee Task Group No. 132: Report.
      ] report, published by the Therapy Physics Committee of the American Association of Physicists in Medicine (AAPM), the role of image registration in treatment planning is discussed and guidelines to perform quality assurance for deformable and rigid image registration are provided. In particular, site-specific recommendations for performing qualitative assessments, and to ensure the clinical acceptability of the registration in routine clinical practice are supplied, since there is no comprehensive ground truth when dealing with DIR in patients. The tolerances indicated for DIR were 1–2 voxels for relevant structures when planning, and visible boundaries within PTV and OAR’s margins during the treatment; for digital phantom test cases 95.0% of voxels has to be within the phantom within 2 mm and the maximum error must be <5.0 mm.
      The use of DIR for CBCT dose calculation was investigated on 202 patient cases (Table 4a) and on deformable phantoms (Table 4b).
      Many different DIR algorithms are available and their accuracy has been discussed in [
      • Song G.
      • Han J.
      • Zhao Y.
      • Wang Z.
      • Du H.
      A Review on Medical Image Registration as an Optimization Problem.
      ,

      Oh S, Jong, Kim S, Deformable image registration in radiation therapy, Radiat Oncol J, 35(2), 2017, pp. 101–111, doi: 10.3857/roj.2017.00325.

      ]. BSpline and Morphons algorithms were observed to be the most commonly used both for patient and phantoms studies, but BSpline has been shown to perform better on patients, independent of the delivery technique [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ,

      Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

      ,
      • Onozato Y.
      • et al.
      Evaluation of on-board kV cone beam computed tomography-based dose calculation with deformable image registration using hounsfield unit modifications.
      ,
      • Veiga C.
      • et al.
      Toward adaptive radiotherapy for head and neck patients: Feasibility study on using CT-to-CBCT deformable registration for ‘dose of the day’ calculations.
      ,
      • Cole A.J.
      • Veiga C.
      • Johnson U.
      • D’Souza D.
      • Lalli N.K.
      • McClelland J.R.
      Toward adaptive radiotherapy for lung patients: Feasibility study on deforming planning CT to CBCT to assess the impact of anatomical changes on dosimetry.
      ,
      • Yang Y.
      • Schreibmann E.
      • Li T.
      • Wang C.
      • Xing L.
      Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation.
      ]. Improved results were reported for prostate and head and neck cancer, compared to lung and liver [

      Marchant TE, Joshi KD, Moore CJ, Accuracy of radiotherapy dose calculations based on cone-beam CT: Comparison of deformable registration and image correction based methods, Phys Med Biol, 63(6), 2018, p. aab0f0, doi: 10.1088/1361-6560/aab0f0.

      ,
      • Giacometti V.
      • et al.
      An evaluation of techniques for dose calculation on cone beam computed tomography.
      ,
      • Huang P.
      • et al.
      Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone-beam CT.
      ,
      • Yang Y.
      • Schreibmann E.
      • Li T.
      • Wang C.
      • Xing L.
      Evaluation of on-board kV cone beam CT (CBCT)-based dose calculation.
      ]. The large mean dose difference reported by Huang et al. [
      • Huang P.
      • et al.
      Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone-beam CT.
      ] between the dose calculated in the liver in week 1 and week 3 was explained by the anatomical changes which occurred during the treatment; the authors suggested prompt adaptive re-planning to avoid liver overdose. Veiga et al. [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ,
      • Veiga C.
      • et al.
      Toward adaptive radiotherapy for head and neck patients: Feasibility study on using CT-to-CBCT deformable registration for ‘dose of the day’ calculations.
      ] used the re-plan rCT (acquired approximately one week before the treatment) to evaluate the accuracy of recalculating the dose using CBCT images to avoid dosimetric errors due to the time gap between pCT and CBCT acquisitions (mean value for dose difference as percentage of the prescribed dose<0.7% for photons, and<2.6% for protons).
      No trend can be observed between different CBCT manufacturers/models (described in Appendix 1) within this review.
      For proton therapy, the most accurate result (dose difference normalized to the maximum dose of −0.5% ± 0.9%) was reported in the only publication retrieved from the literature using a gantry-mounted CBCT clinical system for protons [
      • Veiga C.
      • et al.
      A comprehensive evaluation of the accuracy of CBCT and deformable registration based dose calculation in lung proton therapy.
      ]. All the other publications on proton therapy CBCT dose calculation used CBCT datasets acquired on linacs, where DIR proved to be a valid technique for CBCT dose recalculation [
      • Veiga C.
      • et al.
      Cone-Beam Computed Tomography and Deformable Registration-Based ‘Dose of the Day’ Calculations for Adaptive Proton Therapy.
      ,
      • Veiga C.
      • et al.
      A comprehensive evaluation of the accuracy of CBCT and deformable registration based dose calculation in lung proton therapy.
      ,
      • Arai K.
      • et al.
      Feasibility of CBCT-based proton dose calculation using a histogram-matching algorithm in proton beam therapy.
      ,
      • Park Y.K.
      • Sharp G.C.
      • Phillips J.
      • Winey B.A.
      Proton dose calculation on scatter-corrected CBCT image: Feasibility study for adaptive proton therapy.