Measuring dose in lung identifies peripheral tumour dose inaccuracy in SBRT audit

Purpose: Stereotactic Body Radiotherapy (SBRT) for lung tumours has become a mainstay of clinical practice worldwide. Measurements in anthropomorphic phantoms enable verification of patient dose in clinically realistic scenarios. Correction factors for reporting dose to the tissue equivalent materials in a lung phantom are presented in the context of a national dosimetry audit for SBRT. Analysis of dosimetry audit results is performed showing inaccuracies of common dose calculation algorithms in soft tissue lung target, inhale lung material and at tissue interfaces. Methods: Monte Carlo based simulation of correction factors for detectors in non-water tissue was performed for the soft tissue lung target and inhale lung materials of a modified CIRS SBRT thorax phantom. The corrections were determined for Gafchromic EBT3 Film and PTW 60019 microDiamond detectors used for measurements of 168 SBRT lung plans in an end-to-end dosimetry audit. Corrections were derived for dose to medium (D m,m ) and dose to water (D w,w ) scenarios. Results: Correction factors were up to (cid:0) 3.4% and 9.2% for in field and out of field lung respectively. Overall, application of the correction factors improved the measurement-to-plan dose discrepancy. For the soft tissue lung target, agreement between planned and measured dose was within average of 3% for both film and micro-Diamond measurements. Conclusions: The correction factors developed for this work are provided for clinical users to apply to commissioning measurements using a commercially available thorax phantom where inhomogeneity is present. The end-to-end dosimetry audit demonstrates dose calculation algorithms can underestimate dose at lung tumour/lung tissue interfaces by an average of 2 – 5%.


Introduction
Leveraging improvements in modern radiotherapy equipment and practices, Stereotactic Body Radiotherapy (SBRT) for lung tumours has increasingly become standard practice worldwide [1][2][3][4][5].SBRT offers significant advantages to patients and clinicians; a highly conformal dose delivered to small tumours with increased precision and in a shorter timeframe, all with potential improvements in local control [6][7][8][9].Calculation of delivered dose is difficult in these inhomogeneous and small field conditions, requiring careful verification of the treatment planning systems (TPS) accuracy for optimal patient treatment.Measurements should be performed in realistic phantoms which include low density lung materials in order to mimic the physical radiation interactions that occur in patient lung SBRT.Simple homogenous phantoms commonly used for routine quality assurance are insufficient.
Modern TPS algorithms have been widely studied for their dose calculation accuracy, in particular for heterogenous media, at tissue interfaces and for small field deliveries.Previously reported dosimetry audits and phantom studies have shown discrepancies in TPS calculated dose in tumour and lung materials [10][11][12][13][14][15][16].Lung SBRT dosimetry audits by Lambrecht et al [17] and Distefano et al [15] have utilised the anthropomorphic thorax phantom for verification of lung SBRT treatment deliveries in national audit programs.Iterations of the commercially available CIRS thorax phantom (Norfolk, VA, USA) remain a popular choice for clinics and dosimetry auditing bodies performing measurements of dose in lung and lung tumours [10,[13][14][15]17,18].None of these studies have considered correction factors for reporting dose to the low density lung materials in which the measurements have been performed.The aims of this study are twofold; to present a method for measuring dose in lung and lung tumour equivalent materials, and to evaluate the application of the methodology in the context of an end-to-end SBRT dosimetry audit.This manuscript describes Monte Carlo simulations which consider both the calibration and measurement conditions of radiation detectors to derive corrections for measurements in commercially available lung phantom materials.These corrections can be applied for both dosimetry auditing purposes or more broadly for commissioning of TPS for lung treatments in the clinic.The analysis performed in this study evaluates the implementation of the corrections in film and point dose measurements, and compares the four most common TPS calculation algorithms for dosimetric accuracy in soft tissue lung target and inhale lung materials.
The nomenclature for defining dose calculation methods in use by the TPS; dose to medium, in medium (D m,m ), dose to water, in water (D w, w ) and dose to water, in medium (D w,m ) have been previously defined by Kry et al. [19].
In order to assess the various dose calculation methods employed by the different treatment planning systems we propose three measurement standards to which the TPS dose is compared.We ued the same nomenclature to define our corrections to the measurement: 1. Uncorrected dosimeters which are calibrated to provide dose measurements in water (no correction) 2. Correction to measured dose using a model where inhomogeneities were treated as variable density water (k med , Dw,w ) 3. Correction to measured dose using a model where inhomogeneities were treated as lung tissue (k med , Dm,m ) No correction has been defined for the comparison to D w,m algorithms.
We propose the correction, which is based on the tissue model (standard 3), should give the best assessment of true dose in a patient.Each TPS employs a unique model for calculating dose, and each implementation of the planning algorithm, at the audited centres, could also introduce dosimetric offsets.Therefore, for the purpose of this study we assume that standard 3 represents ground truth.We have not attempted to corroborate this ground truth against any other dosimetric standard.This would require another independent audit service performing the same study on a similar cohort.The presence of standard 2 provides examination of dose to water TPS algorithms in isolation.The uncorrected model (standard 1) is also provided to give historical context, because previous literature have used uncorrected measurements.The relative dose discrepancy between planned and measured dose for each planning system can only be assessed when a uniform standard is applied to all measurements.

Background
Previously published work by the authors details the requirement for algorithm specific correction factors when reporting measured dose in patient equivalent materials [20].The detectors used in the SBRT lung dosimetry audit were calibrated in terms of absorbed dose to water, however were used to measure dose when inserted into inhale lung and soft tissue lung target materials, thus requiring a material conversion correction.
To determine the corrections for each material and reporting mode, four scenarios were simulated as per equations ( 1) & (2); dose in lung (D lung ) or dose in variable density water, in water (D VDw,w ), dose to water in water (D w,w ), dose to the detector in water (D det,w ) and dose to the detector in lung (D det, lung ).Correction factors (k med ) for measurement in CIRS inhale lung and soft tissue lung target materials were determined for both radiochromic film and microDiamond detectors, and the corresponding dose measurements calculated according to equations (3) and (4) [20] using the value of the measured signal from the detector, M, and the calibration for an in-water dosimeter N D,w .Corrections for the inhale lung materials were calculated for in-field and out-of-field voxels.

Medium dependent correction factors (Monte Carlo modelling)
Monte Carlo simulations were performed using the EGSnrc user code DOSXYZnrc [21] for Gafchromic EBT3 measurements (Ashland, Bridgwater, NJ USA).The global electron and photon energies (ECUT and PCUT) were 0.521 MeV and 0.01 MeV respectively.The number of histories simulated was chosen to give a combined overall uncertainty of 0.6 -1.0% per voxel.A 5 × 5 cm field, 6 MV and 10 MV spectra from a previously validated Elekta Synergy Linear Accelerator [22] was used to model the incident beam at 80 cm SSD.This was simulated in the BEAMnrc user code of the EGSnrc [21] library.
EBT3 Gafchromic film consists of 28 µm active layer, between two polystyrene layers of 125 µm [23].This structure was modelled in the coronal plane in the centre of a 20 × 20 × mm 3 'CIRS soft tissue lung target' cube (density 1.055 g/cc).Surrounding the soft tissue lung target cube was an 80 × 80 × 80 mm 3 shell of CIRS inhale lung material (density 0.205 g/cc).This geometry is a slightly simplified representation of the lung tumour and lung material in the phantom.The film, soft tissue target and lung cubes were placed at the centre of a 300 × 300 × 300 mm 3 cube of water.The resolution of the central 70 mm of the whole simulation was modelled in 1.0 mm voxels to evaluate the interface effects, with the surrounding inhale lung voxels at 5 mm, and surrounding water voxels at 100 mm.The density and material of the modelled cubes were varied according to the required calculation methods.Correction factors were established for both in field and out of field regions of the inhale lung material.
The PTW 60019 microDiamond detector (PTW Freiburg, Germany) was simplistically modelled based on manufacturer provided specifications [24], with the stem of the detector as a 6.9 × 6.9 × 15.7 mm 3 cuboid of RW3 (polystyrene521icru, density 1.045 g/cc), and the active diamond layer as a 3.1 × 3.1 × 0.3 mm 3 cuboid (density 3.53 g/cc), located 1 mm from the end of the stem.Surrounding the microDiamond was a 20 × 20 × 20 mm 3 cube of CIRS soft tissue lung target material, and 50 × 50 × 50 mm 3 cube of CIRS inhale lung, replicating the measurement point of the detector in the SBRT phantom.The microDiamond lung cubes were placed at the centre of a 300 × 300 × 300 mm 3 cube of water.The density and material of the modelled cubes were varied according to the required calculation methods.This is a somewhat simplified representation of the microDiamond.
Based on manufacturer supplied material composition data [25], density correction files for the CIRS soft tissue lung target and inhale lung materials were created using the ESTAR program [26].These density correction files were used to generate the pegs4 data for the simulations.For D lung simulations, the mass density of the CIRS soft tissue lung target and inhale lung materials was 1.055 g/cc and 0.205 g/ cc respectively.For D VDw,w calculations, where the voxels are simulated as water with a density of target/lung, the relative electron density of the CIRS soft tissue lung target and inhale lung was 1.028 g/cc and 0.204 g/cc respectively [25].

Lung SBRT dosimetry audit
The Australian Clinical Dosimetry Service (ACDS) SBRT audit was performed on a customised CIRS thorax phantom.The phantom consisted of an anthropomorphic plastic water body with a 20 mm diameter spherical lung tumour in the phantom's right lung surrounded by inhale lung material (Fig. 1a).The lung tumour comprised of CIRS soft tissue target material and was located 6.5 cm from the phantom midline.The audit was an end-to-end test of a facilities' lung SBRT treatment process, including a CT scan, treatment planning and quality assurance according to local SBRT protocol.Phantom setup and delivery of the treatment was performed by the clinical radiation therapists using the local image guidance protocol.No motion management was required for simulation or delivery.The 20 mm lung tumour was taken as the GTV (Gross Target Volume) and a 5 mm margin was applied for PTV (Planning Target Volume), following the TROG 13.01 SAFRON II clinical trial guidelines for SBRT Lung [27].ACDS staff then attended the clinical sites to perform measurements of the plan using a PTW 60019 microdiamond for point dose measurements in the centre of the target and Gafchromic EBT3 Film in the coronal plane, 1 mm anterior to the centre of the target, so that the measured film plane included soft tissue target material surrounded by inhale lung (Fig. 1b).Calibration factors, N D,w , for both detectors were determined by cross calibration to an ARPANSA secondary standard PTW 30013 Farmer chamber.Additional corrections for the microdiamond were required for the nominal small field size and detector orientation [28].
A summary of the 168 plans submitted to the audit are shown in Table 1.Only TPS with five or more plans were included in the analysis.

Application of corrections in dosimetry audit
The 2D correction data from the cube simulation in EGS was adapted to the circular geometry of the audit tumor.This was achieved using an in-house Matlab (Mathworks, Natick MA, USA) program.As shown in Fig. 2a, the cube simulation was reduced to 1D by averaging the horizontal and vertical profiles through the cube target.Once reduced to 1D the simulated data defines corrections in three separate regions: (1) infield-target, (2) in-field-non-target and (3) out-of-field.The interface between in-field and out-of-field zones was defined as where the correction equals 1.000.
The reconstruction of correction data back into the 2D plane of the film is shown in Fig. 2b.The edge-of-field measurement line was defined by the most superior and most inferior points of the 50% isodose line from the uncorrected film measurement.The 1D simulation data, in the superior-inferior dimension, was aligned so that the interface between in-field and out-of-field zones was coincident with edge-of field from the measurement, then copied the across the left-right dimension.For regions in and around the target the 1D target simulation data was copied over a series of rotational transformations into a disk in the coronal plan.A 1 mm Gaussian blur is applied to the 2D correction map to compensate for the blurring in phantom density arising from CT scanning the phantom.
The 2D correction factor maps were applied to each of the film measurements in Matlab according to the primary reporting mode of the algorithm.As detailed in Table 2, the plans submitted in the audit were classified as dose to medium, in medium (D m,m ) or dose to water, in  water (D w,w ).Classification of the Pinnacle CCC algorithm is complex as the algorithm typically reports a mix of D m,m and D w,w .Due to a lack of clarity in the literature [29][30][31][32][33], the audit results for Pinnacle CCC algorithm were analysed using both D m,m and D w,w corrections.
Comparison of the uncorrected and corrected microDiamond and film measurements was made to the planned dose for each audit plan.Dose difference was calculated as (plan-measured)/measured.The results for the four most common TPS algorithms were analysed.The target region in the film was defined as the area bound by the 100% isodose line.The in-field lung region of interest was defined by the area between the 50%-95% isodose.The out of field lung region was defined by the area between the 10%-50% isodose, in all cases excluding the pixels at the interface regions.In order to assess the implementation of the correction factors and avoid dosimetry errors from inaccuracies in the image guided setup, Gafchromic film measurements were aligned to best fit of the planned dose.
Analysis was performed using Matlab and Microsoft Excel.Dose difference was defined as (planned-measured)/measured dose.Dose differences are presented in box and whisker plots showing the mean, median, upper and lower quartiles for each data series.The statistical significance of uncorrected vs corrected results for each algorithm was evaluated using a two sample t-test assuming unequal variances, with the significance level p < 0.05.For the MC simulations, the uncertainty in each voxel was 0.6%-1.0%,leading to a combined overall uncertainty of 0.1-0.2% in the correction factors.

Algorithm specific correction factors
The results of the simulated Gafchromic EBT3 film are shown in Fig. 2a.The lateral central axis correction factor profiles for both the k med D m,m and k med D w,w scenarios are displayed.For the soft tissue lung target and inhale lung (in-field) materials, the difference in correction factors between 6MV and 10MV was <0.6% for both film and micro-Diamond in both k med D m,m and k med D w,w scenarios.For the inhale lung (out-of-field), the differences between 6MV and 10MV were <2.5%.
A summary of the Gafchromic EBT3 film and PTW 60019 micro-Diamond correction factors are also presented in Table 2.The results shown for EBT3 film are for comparison only.For the microDiamond, the correction factors were determined from the voxels in the active layer located 1 mm from the end of the detector.The stated uncertainty only represents the combined statistical uncertainty in the MC simulations.

Audit results
For each audit plan, the differences between planned and measured dose are summarised in Fig. 3, Fig. 4 and Table 3.For the soft tissue lung target measurements, the film and microDiamond agreed to within 1.0% for all algorithms except CCC which showed agreement within 2.5% (Fig. 3a and Fig. 3b).
For both the in-field and out-of-field regions in the inhale lung material, all algorithms showed a reduction in local dose discrepancy with the corrections applied.In the soft tissue target material, for measurements with both film and microDiamond, the corrections reduced the dose discrepancy for all algorithms except MC.On balance, application of the correction factors reduced the measurement-to-plan local dose discrepancy across the four regions of interest in the majority of calculation algorithms.We therefore accept the correction for application in the end-to-end audit.
As the reporting mode of Pinnacle CCC is unclear, any observations cannot be used to assess the performance of the correction.The results show that the D m,m corrections appear to be overcorrecting with respect to the D w,w corrections.This was also observed in the SBRT Spine audit analysis by the authors [20].
When corrected, the microDiamond measurement showed a statistically significant improvement for AXB (p = 0.000), whilst MC showed a statistically significant deterioration (p = 0.000).The difference in the AAA and CCC D w,w results were not statistically significant (p = 0.73 and p = 0.56 respectively).When corrected, the soft tissue lung target film measurements showed statistically significant differences for all algorithms except for AAA (AAA p = 0.80, AXB p = 0.000, MC p = 0.003, CCC D w,w p = 0.000).
For the in-field voxels of the inhale lung material (Fig. 4a), all algorithms underpredicted the dose by an average of 1.7%-3.7%,improving to 1.0-3.0%with correction factors applied.The results were statistically significant for AXB (p = 0.000) and MC (p = 0.006), but not for AAA (p = 0.65) or CCC D w,w (p = 0.37).
In the out of field lung voxels (Fig. 4b), a large range of dose discrepancies was seen across the audit results.On average in the out of field region, all algorithms showed an improvement in results with the application of corrections, however the differences were not statistically significant (AAA p = 0.18, AXB p = 0.08, MC p = 0.14, CCC D w,w p = 0.17).
For each algorithm type, the film results were averaged from all audits to probe the ability of each TPS to accurately model the lung heterogeneity.Fig. 5 shows the 2D average local dose difference maps of all audit films in the scored coronal film plane (<10% dose suppressed) for the four most common TPS algorithms.The 2D maps for the uncorrected measured dose are shown for all algorithms, along with the corrected maps for the primary reporting mode for each algorithm.Also shown is a comparison of the uncorrected and corrected central axis dose profiles, averaged in the centre-superior, centre-inferior, centre-left and centre-right directions.A ring of overdose at the soft tissue lung target/inhale lung interface was seen in all algorithms in the uncorrected dose maps.Application of correction factors to the film improved the average discrepancies across all algorithms.

Discussion
Correction factors are required to accurately measure the dose in and around lung with dosimeters calibrated in water.Location and TPS algorithm specific Monte Carlo derived correction factors for film and microDiamond dosimetry measurements have been provided for a typical lung treatment scenario in a modified commercial phantom.Applying these factors has been shown here to improve audit outcomes for a national SBRT lung audit.Whilst ensuring TPS accuracy is imperative for patient treatment according to the prescription, measurements to assess TPS accuracy are non-trivial.The uncertainties due to heterogeneities are not always thoroughly measured in clinical practice because of inherent dosimetry challenges of small fields and inhomogeneities.The corrections provided in this work can be also used in the clinic to accurately measure the dose in soft tissue lung targets, at the periphery, and in out of field normal lung tissue using commercially available phantoms during commissioning of treatment planning beam models.This publication also presents a methodology for readers to perform MC simulations to determine corrections for other detectors and phantoms.
The radiochromic film corrections presented here for the CIRS lung material can be compared to those presented by Charles et al. [34] for Gafchromic film measurements in ICRU44 simulated lung material.In contrast to our findings, Charles et al. corrections were not necessary for measurements in lung material in field sizes of 3 × 3 cm 3 or greater.However, Monte Carlo simulations in both studies showed a decrease in correction factor, as a function of lateral distance from the central axis of the beam axis.The in-lung, out of field, correction factors exhibited very similar behaviour in both simulationsa sharp decrease in correction factor, up to 1.10 for voxels 2 cm outside the field.The simulations by Charles et al. were performed using a different methodology that did not explicitly include calibration conditions and a slab phantom with water and lung equivalent materials, not considering a tumour embedded in the lung material.An additional consideration is that differences between the simulated ICRU44 lung and CIRS inhale lung materials are likely to be in the order of 2% [25].
In the SBRT audit, the PTV consists of the gross tumour volume made of soft tissue lung target material and the surrounding inhale lung material.Our study showed that without corrections to the film measurements, all algorithms appeared to be reporting a lower dose at the tumour periphery.However, with the application of the correction factors this visual "ring of overdose" improved and only AAA and AXB underestimated the dose at the tumour periphery, particularly in the inline direction.Mixed results were seen for the target material, with correction factors improving the results for all algorithms except for Monaco MC.
Once corrected, Monaco MC showed a statistically significant deterioration in the average local dose difference in the region for both film and microDiamond, moving from reporting a lower dose between 0 and 1% to a higher dose between 2 and 3% (film average/median 3.0%/ 2.7%, p = 0.002 and microDiamond average/median 2.5%/2.1%,p = 0.000, respectively).The discrepancy was only seen in the soft tissue target in lung, with good agreement shown in the inhale lung infield and out of field regions (-0.8%, p = 0.006) and 2.6%, p = 0.15, respectively).The observed offset in the soft tissue lung target is comparable to results seen in the ACDS end-to-end 3DCRT modality.Lehmann et al [35] describes the ACDS Level III 3DCRT dosimetry which is performed immediately prior to all SBRT audits.One of the tests in this modality measures the dose from a 10x10 cm 2 open left lateral beam in two points in the centre of the phantom soft tissue, behind the lung inhomogeneity.Interestingly, data from this test measured with a farmer chamber in the downstream solid water from the lung interface also showed a bias in the    same direction, albeit smaller, of 1.0% averaged from 34 radiotherapy departments using Monaco MC.This potential bias could be due to inaccuracies in beam modelling, or due to a fundamental limitations in the Monaco implementation of the fast Monte Carlo algorithm in the secondary build-up region in lung [36,37].
The AAA algorithm was found to underestimate dose in the PTV in the inhale lung material, with audit film measurements showing an average local difference of − 3.0%.The authors had previously identified the same trend in the inhale lung material in an end-to-end audit of 3D conformal radiotherapy [14,35].In the soft tissue lung target material, the average local difference was − 1.4% and − 2.5% for the microdiamond and film respectively.The results for the soft tissue lung target did not show the previously published behaviour which showed AAA overestimated the dose in tissue behind lung for a single 3DCRT.We hypothesis this difference could be due to secondary build up effects in the much smaller soft tissue lung target in the SBRT audit.The results of the study by Vassiliev et al. [38] confirm the SBRT audit results, with their study showing AAA underestimating the dose to a lung tumour and at the periphery by 2-5% compared to Monte Carlo simulations.Acur-osXB has been noted by many investigators as advantageous compared to the AAA algorithm [29,[39][40][41][42][43].Our results confirm this, with average dose discrepancies of 0.3% (local film)/-0.4%(local microdiamond) and 0.8% (local film) in the soft tissue lung target and in-field inhale lung materials respectively.It should be noted that our study examined plans that were relatively large in field size compared to what may be used in SBRT.Öllers et al. [10] found AXB underestimates dose in very small lung tumours, the underestimation increasing with decreasing field size.Our results found Pinnacle CCC algorithm overestimates dose to the soft tissue lung target by an average of 4.2% (local film/1.8%(local microdiamond) when corrected for D m,m and but is closer to measurement when corrected by D w,w with local differences of 1.3% (film) /-1.5% (microdiamond).This is comparable to the results by Kry et al. [43], who showed an average 3.2% overestimation of dose to a slightly larger lung tumour than that included in our study.The MC algorithm was also found to overestimate dose to the soft tissue lung target (average 3.0% (local film)/2.5% (local microDiamond) local dose difference), however showed good agreement in the inhale lung material (average − 0.8% film local dose difference in-field, and average − 1.0% local dose difference out of field).
In this study it was difficult to ascertain if plan parameters had a measurable impact on the audit outcome.The majority of the audit plans (80%) were delivered using a VMAT delivery technique.Only 6% of plans were submitted with a dose calculation grid > 2 mm 3 , and all plans were within the maximum recommended grid size of 3 mm 3 in the SBRT lung protocol on which the audit was based [27,44].In addition, no link was found between MU/modulation factor and audit result.Further work could be undertaken in a more controlled environment such as the clinic to determine the impact of plan parameters on dosimetric accuracy.
This study identified inaccuracies of common TPS algorithms at predicting the dose in soft tissue lung target and inhale lung materials, particularly at the interfaces.Awareness of algorithm limitations should be considered by clinicians when prescribing patient treatment for lung SBRT.A limitation of this study was delivery of the plan to a static phantom, which does not always reflect the clinical scenario as a patient would usually be treated free breathing or breath hold, both of which would lead to some degree of blurring at the tissue interfaces.In addition, lung tumours in a patient may change shape during treatment.Further work incorporating motion and adaptation should be conducted to better approximate the full complexity of the interplay effect in patient treatment.

Conclusion
The end-to-end audit of lung SBRT demonstrates modern TPS algorithms can accurately predict dose to phantom lung and tumour materials to within 5% as measured by film and microDiamond detectors.Discrepancies between planned and measured dose are seen at the tumour lung/interface, with algorithms underestimating the dose to varying degrees.Correction factors for measuring dose in lung materials of a commercially available thorax phantom are provided for other users to apply in clinical practice.

Fig. 1 .
Fig. 1.(a) Transverse CT view of ACDS SBRT phantom showing Lung_GTV and Lung_PTV structures and (b) photo of coronal plane of the physical film locator insert for lung SBRT target.

Fig. 2 .
Fig. 2. (a) Results of the Monte Carlo simulations (k med ) for CIRS soft tissue lung target and inhale lung (in-and out-of-field) materials, and the corresponding 2D correction factor maps (b) as applied to the audit film geometries using an in-house Matlab program.

Fig. 3 .
Fig. 3. Uncorrected and corrected individual audit results showing local dose difference for microDiamond (a) and film (b) measuring in the soft tissue lung target material.The mean of each data series is indicated by the 'x' on the plots.

Fig. 4 .
Fig. 4. Uncorrected and corrected individual audit results showing local dose difference for film in the in-field (a) and out of field voxels (b) voxels of inhale lung material.The mean of each data series is indicated by the 'x' on the plots.

Fig. 5 .
Fig. 5. Average local dose difference per dose calculation algorithm (a-d) across scored 2D coronal film plane for lung SBRT audit case.Areas of red indicate a higher measured dose relative to the plan, while areas of blue indicate a lower measured dose relative to the plan.A comparison of the uncorrected and corrected dose profiles, averaged in the centre-superior, centre-inferior, centre-left & centre-right, across central axis is also shown.The grey shading indicates the area in the soft tissue lung target material.Analysis is performed with film aligned to best fit with planned dose.

Table 1
Summary of audit plan parameters.

Table 2
Algorithm specific correction factors (k med ) for CIRS soft tissue lung target and inhale lung (in-and out-of-field) materials.

Table 3
Summary of the average local dose differences for the soft tissue lung target and inhale lung audit measurements for uncorrected, D m,m corrected and D w,w corrected as appropriate for each TPS algorithm.