Original paper| Volume 98, P98-112, June 01, 2022

# A novel tool for motion-related dose inaccuracies reduction in 99mTc-MAA SPECT/CT images for SIRT planning

Published:May 05, 2022

## Highlights

• SPECT/CT reconstruction is affected by motion.
• Liver and tumor dosimetry in radioembolization may be improved by motion correction.
• A dedicated tool was implemented to realign the projection images.
• The tool was experimentally validated with an ad-hoc dynamic phantom.
• The correction improved the lung, liver and tumor dose estimation.

## Abstract

### Introduction

In Selective Internal Radiation Therapy (SIRT), 90Y is administered to primary/secondary hepatic lesions. An accurate pre-treatment planning using 99mTc-MAA SPECT/CT allows the assessment of its feasibility and of the activity to be injected. Unfortunately, SPECT/CT suffers from patient-specific respiratory motion which causes artifacts and absorbed dose inaccuracies. In this study, a data-driven solution was developed to correct the respiratory motion.

### Methods

The tool realigns the barycenter of SPECT projection images and shifts them to obtain a fine registration with the attenuation map. The tool was validated using a modified dynamic phantom with several breathing patterns. We compared the absorbed dose distributions derived from uncorrected($Dm$)/corrected($Dc$) images with static ones($Ds$) in terms of γ-passing rates, 210 Gy isodose volumes, dose-volume histograms and percentage differences of mean doses (i.e., $ΔD¯m$ and $ΔD¯c$, respectively). The tool was applied to twelve SIRT patients and the Bland-Altman analysis was performed on mean doses.

### Results

In the phantom study, the agreement between $Dc$ and $Ds$ was higher (γ-passing rates generally > 90%) than $Dm$ and $Ds$. The isodose volumes in $Dc$ were closer than $Dm$ to $Ds$, with differences up to 10% and 30% respectively. A reduction from a median $ΔD¯m$ = -19.3% to $ΔD¯c$ = -0.9%, from $ΔD¯m$ = -42.8% to $ΔD¯c$ = -7.0% and from $ΔD¯m$ = 1586% to $ΔD¯c$ = 47.2% was observed in liver-, tumor- and lungs-like structures. The Bland-Altman analysis on patients showed variations (±50 Gy) and (±4 Gy) between $D¯c$ and $D¯m$ of tumor and lungs, respectively.

### Conclusion

The proposed tool allowed the correction of 99mTc-MAA SPECT/CT images, improving the accuracy of the absorbed dose distribution.

## Keywords

#### Abbreviations:

SIRT (Selective Internal Radiation Therapy), TARE (Transarterial Radioembolization), PLLA (Poly-L-Lactic Acid), CECT (Contrast-Enhanced Computed Tomography), MR (Magnetic Resonance), MAA (Macro-Aggregated Albumin), LSF (Lung Shunt Fraction), VOI (Volume Of Interest), OARs (Organs At Risk), DVH (Dose Volume Histogram), MC (Monte Carlo), μMap (Attenuation Map), PET (Positron Emission Tomography), CIRS (Computerized Imaging Reference System), OSEM (Ordered Subset Expectation Maximization), GUI (Graphical User Interface), LDM (Local Deposition Method), DD (Dose Difference), DTA (Distance To Agreement), LOA (Limits Of Agreement), SPECT (Single Photon Emission Computed Tomography)

## 1. Introduction

Selective internal radiation therapy (SIRT), also known as transarterial radioembolization (TARE), is one of the most effective treatments for unresectable primary and secondary hepatic lesions. SIRT is based on the administration of 90Y resin (SIR-Spheres®, Sirtex Medical Ltd, Woburn, MA, USA) or glass (TheraSphere™, Boston Scientific, Marlborough, MA, USA) or 166Ho poly-L-Lactic Acid (PLLA) (QuiremSpheres®, Quirem Medical BV, Deventer, the Netherlands) microspheres. The recommended treatment planning procedure relies on two important phases: first is the use of high-resolution multi-phase Contrast-Enhanced Computed Tomography (CECT) or Magnetic Resonance (MR) [
• Atassi B.
• Bangash A.K.
• Bahrani A.
• Pizzi G.
• Lewandowski R.J.
• Ryu R.K.
• et al.
Multimodality Imaging Following 90Y Radioembolization: A Comprehensive Review and Pictorial Essay.
] for appropriate target definition; second is the use of 99mTc macro-aggregated albumin (99mTc-MAA) planar or Single Photon Emission Computed Tomography (SPECT)/CT imaging as a surrogate of the treatment device distribution. More recently scout dose of 166Ho-PLLA has been approved for planning SIRT treatments, allowing to use of the same radioactive device for pre and post-treatment imaging [
• Smits M.L.J.
• Dassen M.G.
• Prince J.F.
• Braat A.J.A.T.
• Beijst C.
• Bruijnen R.C.G.
• et al.
The superior predictive value of (166)Ho-scout compared with (99m)Tc-macroaggregated albumin prior to (166)Ho-microspheres radioembolization in patients with liver metastases.
,
• Chiesa C.
• Maccauro M.
(166)Ho microsphere scout dose for more accurate radioembolization treatment planning.
].
The pre-treatment simulation is mandatory to evaluate the possible SIRT extrahepatic shunts [
• Cremonesi M.
• Chiesa C.
• Strigari L.
• Ferrari M.
• Botta F.
• Guerriero F.
• et al.
]. A Lung Shunt Fraction (LSF) >20% or any extrahepatic shunt are exclusion criteria for SIRT patients. Thus, this quantity needs an accurate calculation to prevent the exclusion of patients potentially benefiting from radioembolization. Notably, the MAA presents a variety in size (0 to 150 μm) with respect to the resin microspheres (20 to 60 μm); thus, 99mTc-MAA images may overestimate the LSF, especially when planar scintigraphy imaging is used [
• Elsayed M.
• Cheng B.
• Xing M.
• Sethi I.
• Brandon D.
• Schuster D.M.
• et al.
Comparison of Tc-99m MAA Planar Versus SPECT/CT Imaging for Lung Shunt Fraction Evaluation Prior to Y-90 Radioembolization: Are We Overestimating Lung Shunt Fraction?.
,
• Chiesa C.
• Lambert B.
• Maccauro M.
• Ezziddin S.
• Dieudonné A.
• et al.
Pretreatment Dosimetry in HCC Radioembolization with (90)Y Glass Microspheres Cannot Be Invalidated with a Bare Visual Evaluation of (99m)Tc-MAA Uptake of Colorectal Metastases Treated with Resin Microspheres.
,
• Ulrich G.
• Dudeck O.
• Furth C.
• Ruf J.
• Grosser O.S.
• et al.
Predictive value of intratumoral 99mTc-macroaggregated albumin uptake in patients with colorectal liver metastases scheduled for radioembolization with 90Y-microspheres.
] without the attenuation correction.
Although SPECT/CT has been proposed for LSF estimation to overcome the limitation of planar imaging [
• Allred J.D.
• Niedbala J.
• Mikell J.K.
• Owen D.
• Frey K.A.
• Dewaraja Y.K.
The value of (99m)Tc-MAA SPECT/CT for lung shunt estimation in (90)Y radioembolization: a phantom and patient study.
], patient respiratory motion during the SPECT/CT acquisition still causes a blurring of the activity distribution, which further affects the LFS estimation especially when the target lesions are placed a few centimeters from the hepatic dome. To improve the LSF estimation accuracy, several authors proposed an ad-hoc reduction of Volume of Interest (VOI) by removing the cranial lungs contours placed within an empirically determined distance (1.5 or 2 cm) from the apex of both abdominal diaphragms [
• Kao Y.H.
• Magsombol B.M.
• Toh Y.
• Tay K.H.
• Chow P.KH.
• Goh A.SW.
• et al.
Personalized predictive lung dosimetry by technetium-99m macroaggregated albumin SPECT/CT for yttrium-90 radioembolization.
,
• Yu N.
• Srinivas S.M.
• Difilippo F.P.
• Shrikanthan S.
• Levitin A.
• McLennan G.
• et al.
Lung dose calculation with SPECT/CT for 90Yittrium radioembolization of liver cancer.
]. This approach was also validated using a static phantom, accounting for the scatter radiation but not considering the respiratory movement [
• Allred J.D.
• Niedbala J.
• Mikell J.K.
• Owen D.
• Frey K.A.
• Dewaraja Y.K.
The value of (99m)Tc-MAA SPECT/CT for lung shunt estimation in (90)Y radioembolization: a phantom and patient study.
].
In addition to LSF estimation, pre-therapy SPECT/CT imaging is essential for an accurate estimation of the activity distribution in tumors and organs at risk (OARs) in order to obtain an accurate image-based dosimetry [
• Chiesa C.
• Sjogreen-Gleisner K.
• Walrand S.
• Strigari L.
• Flux G.
• Gear J.
• et al.
EANM dosimetry committee series on standard operational procedures: a unified methodology for (99m)Tc-MAA pre- and (90)Y peri-therapy dosimetry in liver radioembolization with (90)Y microspheres.
,
• Dewaraja Y.K.
• Frey E.C.
• Sgouros G.
• Brill A.B.
• Roberson P.
• Zanzonico P.B.
• et al.
MIRD pamphlet No. 23: quantitative SPECT for patient-specific 3-dimensional dosimetry in internal radionuclide therapy.
,
• Chiesa C.
• Strigari L.
• Pacilio M.
• Richetta E.
• Cannatà V.
• Stasi M.
• et al.
Dosimetric optimization of nuclear medicine therapy based on the Council Directive 2013/59/EURATOM and the Italian law N. 101/2020. Position paper and recommendations by the Italian National Associations of Medical Physics (AIFM) and Nuclear Medicine (AIMN).
], which can be used to determine the optimal activity to be injected for the SIRT approach [
• Walrand S.
• Hesse M.
• Amini N.
• Lhommel R.
• et al.
Accurate non-tumoral 99mTc-MAA absorbed dose prediction to plan optimized activities in liver radioembolization using resin microspheres.
,
• Garin E.
• Tselikas L.
• Guiu B.
• Chalaye J.
• Edeline J.
• de Baere T.
• et al.
Personalised versus standard dosimetry approach of selective internal radiation therapy in patients with locally advanced hepatocellular carcinoma (DOSISPHERE-01): a randomised, multicentre, open-label phase 2 trial.
] so that it is possible to retrospectively evaluate the absorbed-dose effect relationships [
• Strigari L.
• Sciuto R.
• Rea S.
• Carpanese L.
• Pizzi G.
• Soriani A.
• et al.
Efficacy and toxicity related to treatment of hepatocellular carcinoma with 90Y-SIR spheres: radiobiologic considerations.
] or to assess the agreement between the distribution of 99mTc-MAA and 90Y [
• Bernard C.
• Lovinfosse P.
• Gérard L.
• Lilet H.
• Cornet O.
• et al.
Comparative dosimetry between (99m)Tc-MAA SPECT/CT and (90)Y PET/CT in primary and metastatic liver tumors.
,
• Kafrouni M.
• Allimant C.
• Vauclin S.
• Guiu B.
• Mariano-Goulart D.
• et al.
Analysis of differences between (99m)Tc-MAA SPECT- and (90)Y-microsphere PET-based dosimetry for hepatocellular carcinoma selective internal radiation therapy.
].
Image-based dosimetry can be performed with the partition model, which allows to estimate the mean absorbed dose to the partitioned liver, or with voxel-based dosimetry, which allows estimation of the dose-volume histograms (DVHs) while considering the inhomogeneity of the absorbed dose distribution inside each compartment [
• Kim S.P.
• Cohalan C.
• Kopek N.
• Enger S.A.
A guide to (90)Y radioembolization and its dosimetry.
].
In voxel-based dosimetry, in particular, the respiratory motion affects the DVHs calculation as shown by Bastianeet et al. [
• Bastiaannet R.
• Viergever M.A.
• de Jong H.W.A.M.
Impact of respiratory motion and acquisition settings on SPECT liver dosimetry for radioembolization.
] who systematically evaluated the combined contribution of simulated breathing, patient-based characteristics (i.e., sex, weight, tumor size and location), and acquisition parameters in digital experiments. More in detail, the authors assessed DVHs obtained by synthetic SPECT/CT images generated with a Monte Carlo (MC) simulation and reproducing patient-like virtual phantoms with multiple liver cancer lesions in breath-hold or free-breathing conditions.
For all of these reasons, an approach to correct the artifacts generated by the respiratory motion is needed for accurate pre-treatment dosimetry.
Motion correction in SPECT/CT imaging has been widely investigated in the context of cardiac imaging [
• Bitarafan-Rajabi A.
• Rajabi H.
• Rastgou F.
• Yaghoobi N.
• Malek H.
• et al.
Influence of respiratory motion correction on quantification of myocardial perfusion SPECT.
,
• Dasari P.K.R.
• Könik A.
• Pretorius P.H.
• Johnson K.L.
• Segars W.P.
• Shazeeb M.S.
• et al.
Correction of hysteretic respiratory motion in SPECT myocardial perfusion imaging: Simulation and patient studies.
,
• Daou D.
• Sabbah R.
• Coaguila C.
• Boulahdour H.
Feasibility of data-driven cardiac respiratory motion correction of myocardial perfusion CZT SPECT: A pilot study.
,
• Kovalski G.
• Israel O.
• Keidar Z.
• Frenkel A.
• Sachs J.
• Azhari H.
Correction of heart motion due to respiration in clinical myocardial perfusion SPECT scans using respiratory gating.
,
• Polycarpou I.
• Chrysanthou-Baustert I.
• Parpottas Y.
• Panagidis C.
• Marsden P.K.
• et al.
Impact of respiratory motion correction on SPECT myocardial perfusion imaging using a mechanically moving phantom assembly with variable cardiac defects.
,
• Segars W.P.
• Mok S.P.
• Tsui B.M.W.
Investigation of Respiratory Gating in Quantitative Myocardial SPECT.
,

Healthcare G. Motion Detection and Correction (MDC) on Xeleris: white paper.

], although these methods generally need external devices or dedicated SPECT/CT systems. Moreover, data driven solutions have been proposed for motion correction in LSF estimation with SPECT imaging and radioembolization dosimetry [
• Sanders J.C.
• Ritt P.
• Kuwert T.
• Vija A.H.
• Maier A.K.
Fully Automated Data-Driven Respiratory Signal Extraction From SPECT Images Using Laplacian Eigenmaps.
]. In the work originally proposed by Sanders et al. [
• Sanders J.C.
• Ritt P.
• Kuwert T.
• Vija A.H.
• Maier A.K.
Fully Automated Data-Driven Respiratory Signal Extraction From SPECT Images Using Laplacian Eigenmaps.
] and furtherly investigated by Robert et al. [
• Robert A.
• Rit S.
• Baudier T.
• Jomier J.
• Sarrut D.
Data-Driven Respiration-Gated SPECT for Liver Radioembolization.
], the projection images in raw-data format were acquired at high frame-rate (5 Hz) during the SPECT acquisition, grouped in 10 respiratory phases using the Laplacian eigenmaps and finally reconstructed to generate a 4D SPECT image. Nevertheless, the feasibility of this approach strictly depends upon the availability of the list-mode for projection images acquisition, which is not available in conventional SPECT/CT systems [
• Ritt P.
Recent Developments in SPECT/CT.
], and could be limited by a lower count statistic if compared with PET systems [
• Sanders J.C.
• Ritt P.
• Kuwert T.
• Vija A.H.
• Maier A.K.
Fully Automated Data-Driven Respiratory Signal Extraction From SPECT Images Using Laplacian Eigenmaps.
,
• van der Vos C.S.
• Koopman D.
• Rijnsdorp S.
• Arends A.J.
• Boellaard R.
• van Dalen J.A.
• et al.
Quantification, improvement, and harmonization of small lesion detection with state-of-the-art PET.
].
Dietze et al. [
• Dietze M.M.A.
• Bastiaannet R.
• Kunnen B.
• Velden S.
• Lam M.G.E.H.
• Viergever M.A.
• et al.
Respiratory motion compensation in interventional liver SPECT using simultaneous fluoroscopic and nuclear imaging.
] proposed another data driven method for motion compensation in liver radioembolization which was based on the tracking of the centre of mass, extracted from fluoroscopic projections and used to reconstruct the motion signal. In this study, the amplitudes of the hepatic dome movements were derived from fluoroscopic images and applied to simultaneously collected SPECT images. This approach led to the development of a new hardware and software solution to correct SPECT and interventional images from motion artifacts [
• Dietze M.M.A.
• Kunnen B.
• Brontsema F.
• Ramaekers P.
• Beijst C.
• Afifah M.
• et al.
A compact and mobile hybrid C-arm scanner for simultaneous nuclear and fluoroscopic image guidance.
]. This new device was tested using a digital [
• Dietze M.M.A.
• Bastiaannet R.
• Kunnen B.
• Velden S.
• Lam M.G.E.H.
• Viergever M.A.
• et al.
Respiratory motion compensation in interventional liver SPECT using simultaneous fluoroscopic and nuclear imaging.
] or an anthropomorphic phantom [
• Dietze M.M.A.
• Kunnen B.
• Lam M.G.E.H.
• de Jong H.W.A.M.
Interventional respiratory motion compensation by simultaneous fluoroscopic and nuclear imaging: a phantom study.
] placed on a motor-driven stage. Unfortunately, this tool requires a dedicated software/hardware which is currently unavailable in the clinical practice.
In addition to the critical issues already mentioned, all the above described approaches were focused on artifact-free SPECT reconstruction but did not take into account its registration with the attenuation map (µMap) for hepatic lesions. The µMap is generated from a free breathing CT, acquired in a SPECT/CT system, and its registration with the SPECT is crucial for a proper attenuation correction and an accurate quantification of LSF, as well as tumor and OARs dosimetry.
To improve the pre-treatment accuracy of 99mTc-MAA SPECT/CT-derived absorbed dose distribution, we propose an alternative data-driven solution, in principle applicable to all the commercial SPECT/CT systems.
This approach simultaneously allows for the realignment of the SPECT projection images and subsequent shifting to obtain a finer registration between SPECT and CT. The former is achieved via the calculation of baricenter of collected counts on the tumoral area, the latter relies on the extraction od the information on image displacement from the µMap.
To validate our method a commercially available moving phantom was modified with a novel ad-hoc solution to simulate liver tumors located close to the hepatic dome. Finally, we applied the developed correction approach to a patient cohort and investigated the possible improvement of the target/OARs dose distribution estimation and their correlation with the patient follow-up images.

## 2. Methods

### 2.1 Phantom study

#### 2.1.1 Phantom description

In this study we propose an ad-hoc phantom which simulates the respiratory motion of a patient having a spherical tumoral lesion and is suitable for SPECT/CT or Positron Emission Tomography(PET)/CT acquisition images.
To do so, a Computerized Imaging Reference Systems (CIRS, Norfolk, VA) Dynamic Thorax phantom model 008A, designed for gated radiotherapy treatment, was adapted with a 3D-printed component composed of a hollow sphere (diameter of 50 mm, internal volume of about 65.5 ml) connected to the moving system of the CIRS phantom through a cylindrical bar of about 20 cm. The sphere was filled with 174 MBq of 99mTc, obtained as sodium pertechnetate (Na[99mTc]O4) from 99Mo/99mTc generator (Ultratechnekow, CURIUM, Netherlands), and was realised to mimic a moving spherical tumor nearby the hepatic dome. A cardboard box shaped as the CIRS phantom and containing a plastic tube to ease the sphere movement surrounded by water bags was connected to the end of the phantom to simulate the patient abdomen.
We used this phantom to validate the GUI described in the paragraph 2.4.

#### 2.1.2 Investigated respiratory cycles for phantom validation

The bar with the radioactive-filled sphere was controlled by the CIRS motion control software which includes the Trio PC Motion library to generate several ad hoc sinusoidal movements mimicking the excursion of a lesion located close to the hepatic dome. The amplitudes and the cycle durations of the sphere movement were derived from [
• Keall P.J.
• Mageras G.S.
• Balter J.M.
• Emery R.S.
• Forster K.M.
• Jiang S.B.
• et al.
The management of respiratory motion in radiation oncology report of AAPM Task Group 76a).
,
• Yuan G.
• Drost N.A.
• McIvor R.A.
Respiratory rate and breathing pattern.
], respectively. Fig. 1 shows the experimental setup. The phantom was acquired in 10 experimental setups (1 static and 9 dynamic setups). Among these, acquisitions #2, #3, #4, #7, and #8 represented normal breathing types; acquisitions #5 and #6 were considered as deep breathing (i.e., amplitudes of 20 and 25 mm, respectively), while acquisitions #9 and #10 were set up as abnormally prolonged breathing cycle durations (12 and 40 s, respectively).
All the dynamic setups followed a sinusoidal wave in the $z$ direction with the amplitudes and cycle durations specified in Table 1. After the acquisition, the SPECT projection images of the moving phantom were corrected using the ad-hoc MATLAB (R2020b, The MathWorks Inc., Natick, MA, USA) tool.
Table 1Breathing type (i.e. normal, abnormal, deep inspiration and static), cycle duration, number of breathing per minute, the maximum amplitude of sphere movement in the z direction (CC) and the total movement.
Acquisition #Breathing typeCycle duration [s]Number of breathing per minute [min-1]Maximal amplitude (in the CC direction) [mm]Total movement (CC direction) [mm]
1Static0000
2Normal512510
3Normal5127.515
4Normal51212.525
5Deep5122040
6Deep5122550
7Normal32012.525
8Normal41512.525
9Abnormal12512.525
10Abnormal401.512.525

#### 2.1.3 Phantom contouring

Liver-, lungs- and tumor-like structures were drawn on the SPECT/CT images of the static acquisition using the MIM Maestro software (MIM Software Inc., Cleveland, OH). The lungs- and tumor-like VOI were obtained by manually segmenting the lungs of CIRS dynamic phantom and the sphere on the CT of the SPECT/CT image, respectively. The liver-like VOI was obtained by the rigid co-registration of the phantom CT image with that of an example liver cancer patient. The liver-like VOI was positioned adjacent to the end of the the lungs-like VOI in the z (i.e., cranio-caudal) direction. In addition, the most inferior 2 cm-thick volumes of the left and right lungs were automatically subtracted from the lungs-like VOIs (lungs-2 cm-like) as described in [
• Allred J.D.
• Niedbala J.
• Mikell J.K.
• Owen D.
• Frey K.A.
• Dewaraja Y.K.
The value of (99m)Tc-MAA SPECT/CT for lung shunt estimation in (90)Y radioembolization: a phantom and patient study.
].

### 2.2 Patients study

#### 2.2.1 Patients cohort description

After the phantom validation phase, 12 consecutive single-lesion liver cancer patients who received 90Y radioembolization treatment were enrolled in this study to assess the impact of the proposed correction using the developed tool on the projection images of the pre-treatment 99mTc-MAA SPECT/CT. All patients underwent images as described in paragraph 2.3. Follow-up CECT were used to assess the local response of patients to the SIRT treatment.

#### 2.2.2 Patients contouring

Tumor delineation on patients images was jointly performed by a Radiation Oncology physician and a Radiologist using MIM by reviewing the arterial or venous phases of a multiphasic CECT acquired within one to three weeks from the planning angiographic procedure. This approach was followed to obtain an accurate target delineation that is not feasible on the low-dose CT from the hybrid SPECT/CT system, as it is acquired in free-breathing condition without contrast media. This CT image will be referred to as CT in the following sections to distinguish it from the CECT.
Liver and lungs contours were delineated by a Radiation Oncology physician on the CT with manual and semi-automated MIM tools. As in the phantom study, a lungs-2 cm VOI was delineated for each patient. The CECT was registered to the CT focusing on the liver to maintain the relative distance between the tumor and the hepatic dome and the tumor contour was rigidly transferred from the CECT to the CT.

### 2.3 Image acquisition

The SPECT/CT images were acquired on a Discovery 670 NM/CT (GE Healthcare, Milwaukee, USA) with the standard 99mTc-MAA protocol (acquired using LEHR collimator, step-and-shoot technique, 15 s/view, 3° angular step, 128x128 matrix size per projection, emissive window centred at 140.5 keV with a width of 10 keV and scatter window centred at 120 keV with a width of 5 keV).
All the SPECT images were reconstructed using the ordered subset expectation maximization (OSEM) algorithm with 2 iterations and 10 subsets, the attenuation (based on the CT) and scatter correction, the resolution recovery and no post-processing filter. All the CT were acquired in helical mode with a rotation time of 0.8 s, slice thickness of 3.75 mm, pitch of 1.35, large SFOV, 120 kV, 80 mA and reconstructed with GE Standard (STD) algorithm.

### 2.4 Developed image processing tool

The tool, developed with MATLAB, is structured as a graphical user interface (GUI) and its workflow is described in Fig. 2 and can be summarized in realignment of the projection images and shifting between the projection images and the µMap. The MATLAB GUI layout for patients is showed in Fig. 3.
The workflow (Fig. 2) requires as input the CT, the µMap generated by the dual-head SPECT/CT system, the 2D projection images (pj,k, where j and k are the indexes of the projection images, ranging from 1 to 60, and of the SPECT head, respectively) of the SPECT acquisition, and the contours of the tumor and the liver delineated on the CT, which can be uploaded trough the developed GUI (Fig. 3).
A threshold of 800 HU on the CT images allows the automatic generation of the body contour using the regionprops MATLAB function. The calculation tool computes the distance between the tumor isocentre ($xt,yt,zt$) and the body contour along the $x$ ($dxCT$) and $y$ ($dyCT)$ directions, defined as $dxCT=xbody,y=ytCT-xt$ and $dyCT=ybody,x=xtCT-yt$, respectively. This is done to select the closest projection image to the tumor ($pbest,kbest$), which is supposed to be the least affected by the scattering of the patient body and to represent the most accurate 2D representation of the tumor.
The amplitude of the tumor contour along the z (cranio-caudal) direction ($AtCT$) is obtained from the tumor VOI delineated on the CT and a percentage threshold (Th%) is recursively set to match the size of the thresholded area on $pbest,kbest$ along the $z$ direction to $AtCT$.
Th% is then applied to all the projection images $pj,k$ to get the z-coordinate of the geometrical barycentre of the high-counts area (i.e., tumor), called $Btpj,k$ and defined as:
$Btpj,k=∑i=1Mj,kzi,tpj,kMj,k$

Where $zi,tpj,k$ and $Mj,k$ represent the z coordinate ot the i-th pixel and the total number of pixels contained in the thresholded area, respectively. For display purposes, the upper and lower z-coordinate of the high-counts area are also obtained and called $Utpj,k$ and $Ltpj,k$, respectively. The median z-coordinate of the barycentres of projection images belonging to $kbest$, i.e., $median(Btpj,k=kbest)$ is calculated.
In order to match $Btpj,kbest$ to $median(Btpj,k=kbest$), all the projection images are realigned by:
$ΔBtpj,k=Btpj,kbest-medianBtpj,k=kbest$
(2)

Where the assumption $ΔBtpj,k≠kbest=ΔBtpj,k=kbest$ is made since the projection images obtained by the furthest SPECT head (i.e., $k≠kbest$) are characterized by high scattering rate and poor resolution.
In the registration phase, the tool accounts for the correct positioning of the projection images with respect to the µMap.
For this purpose, the distance between the upper point of the tumor and the hepatic dome along the $z$ coordinate (i.e., $ddome-tCT$) is computed from the contours of the tumor and the liver delineated on the CT.
At the same time, the positions of the upper point of the tumor and the hepatic dome are identified on the µMap and $pbest,kbest$, respectively, and their distance (i.e., $ddome-tμMap,pbest,kbest$) is measured. The µMap and $pbest,kbest$ are natively registered, so $ddome-tμMap,pbest,kbest$ is simply obtained as a difference between the $z$ coordinates of the two points. The upper point of the hepatic dome (i.e., $UdomeμMap$) on the µMap is automatically identified by considering 11 sagittal slices located around the mid thickness of the patient. On each sagittal slice a threshold of 150 a.u. is applied to distinguish the hepatic dome by the lung tissue and the sagittal slice containing the highest point of the hepatic dome is retrieved and showed to the user. The upper point of the tumor on $pbest,kbest$ (i.e., $Utpbest,kbest$) is obtained by considering the maximum z coordinate of the thresholded area in $pbest,kbest$, as previously described.
Defined $∊=ddome-tμMap,pbest,kbest-ddome-tCT$, the registration between the µMap and the corrected projection images $pj,k$, is therefore adjusted by shifting them such as:
$∊→0,thusddome-tμMap,pbest,kbest=ddome-tCT$
(3)

The projection images obtained are finally saved and can be imported in a software for the 3D SPECT reconstruction with attenuation correction, scattering correction and resolution recovery settings.
For what concerns the moving phantom acquisitions, the same CT and the CT-derived µMap of the static acquisition were used in the MATLAB tool for all the acquisitions. In addition, the registration-related shifting was performed by considering the slice adjacent to the lowest point of the lungs insert on the CIRS phantom as the upper point of the hepatic dome on the µMap.

### 2.5 Absorbed dose distributions calculation

The SPECT images (using the projection images before and after the application of the MATLAB tool, i.e., uncorrected and corrected, respectively) for both the phantom and patients studies were reconstructed in Xeleris (GE Healthcare, Milwaukee, USA) and subsequently imported in MIM software for the absorbed dose distribution calculation. For the following analyses, the subscripts s, m and c indicate the static, the uncorrected moving and the corrected moving setups, respectively (e.g., SPECTc corresponds to the tool-corrected SPECTm image). Patients images consisted of SPECTm and SPECTc only as images without respiratory motion were not available.
The absorbed dose distributions were calculated by the MIM software using a workflow based on the local deposition method (LDM), assuming to fill the 3D-printed sphere used in the phantom study with 1 GBq of 90Y in a relative calibration approach. For patients study, the clinically administrated activities were used in the absorbed dose calculation, using the same MIM workflow with LDM and relative calibration approach. A local density correction was applied to the lungs and lungs-2 cm VOIs, assuming a uniform density of 0.33 g/cm3, being the LDM workflow based on the assumption of a uniform water-like density of 1 g/cm3.

### 2.6 Evaluation of the absorbed dose distributions

To quantify the agreement between absorbed dose distributions obtained from SPECTm and SPECTc (i.e., $Dm$ and $Dc$, respectively) versus SPECTs (i.e., $Ds$) in the phantom study, the gamma index values [
• Low D.A.
• Harms W.B.
• Mutic S.
• Purdy J.A.
A technique for the quantitative evaluation of dose distributions.
] were calculated. The gamma index method is currently adopted in external beam radiotherapy and quantifies the agreement between an evaluated and a reference absorbed dose distributions by simultaneously considering the percentage dose difference (DD) and the distance to agreement (DTA). The Euclidean distance $Γ(r1,r2)$ between two points in the evaluated and in the reference absorbed dose distributions ($r1$ and $r2$, respectively) is computed in a 4-dimensional space renormalized by the DD and DTA criteria ($δr$ and $δD$, respectively), i.e.:
$Γr1,r2=Δrr1,r22δr2+ΔDr1,r22δD2$

Where $Δrr1,r2$ and $ΔDr1,r2$ are the geometric distance and the absorbed dose difference between $r1$ and $r2$, respectively [
• Hussein M.
• Clark C.H.
• Nisbet A.
Challenges in calculation of the gamma index in radiotherapy - Towards good practice.
].
The gamma index is then taken for each point of the reference absorbed dose distribution as the minimum distance in the renormalized space over all the evaluated points, i.e.,
$γr2=minΓr1,r2∀r1$

$γ$ values between 0 and 1 indicate that the comparison passed with respect to the dose and distance criteria, while values>1 indicate failure [
• Miften M.
• Olch A.
• Mihailidis D.
• Moran J.
• Pawlicki T.
• Molineu A.
• et al.
Tolerance limits and methodologies for IMRT measurement-based verification QA: Recommendations of AAPM Task Group No. 218.
]. The $γ$-passing rate indicates the percentage of points satisfying the condition $γr2<1$. The global normalization was applied, in which the DD is normalized using the same value for all point pairs. The δr and δD criteria were fixed to voxel size (4.42 mm) and 5% of the maximum of the reference absorbed dose distribution, respectively, as in [
• Maughan N.M.
• Garcia‐Ramirez J.
• Arpidone M.
• Swallen A.
• Laforest R.
• Goddu S.M.
• et al.
Validation of post-treatment PET-based dosimetry software for hepatic radioembolization of Yttrium-90 microspheres.
]. A dose threshold of 1 Gy on the reference absorbed dose distribution was set to exclude low-dose areas which had little clinical relevance but could significantly affect the results of the $γ$-analysis.
The gamma index was calculated in different volumes (i.e., liver, lungs and the union of liver and lungs). We assumed 90% and 95% as acceptable and optimal cut-off levels for gamma index, respectively, as reported in the AAPM TG 218 recommendations [
• Miften M.
• Olch A.
• Mihailidis D.
• Moran J.
• Pawlicki T.
• Molineu A.
• et al.
Tolerance limits and methodologies for IMRT measurement-based verification QA: Recommendations of AAPM Task Group No. 218.
]. The $γ$-passing rates were calculated using the MIM software [
• Maughan N.M.
• Garcia‐Ramirez J.
• Arpidone M.
• Swallen A.
• Laforest R.
• Goddu S.M.
• et al.
Validation of post-treatment PET-based dosimetry software for hepatic radioembolization of Yttrium-90 microspheres.
] and reported.
The absorbed dose level that identified an isodose volume corresponding to the physical volume of the 3D-printed sphere (i.e., 65.5 ml) in the reference absorbed dose distribution was extracted. The same absorbed dose level was applied to the dose distributions obtained from SPECTm and SPECTc to quantify the change in volume size.
For both phantom and patients studies, DVHs and mean absorbed doses ($D¯$) of the contoured VOIs were extracted in “csv” format to be analyzed. The $D¯$ were calculated for all VOIs and absorbed dose distributions derived from SPECTm, SPECTc and SPECTs configurations, i.e., $D¯m$, $D¯c$ and $D¯s$, respectively. The differences between $D¯m$ and $D¯c$ compared to the reference dose $D¯s$ are indicated as $ΔD¯m$ or $ΔD¯c$, respectively.
A qualitative comparison of the activity distribution in SPECTm/CT and SPECTc/CT images was showed for an example patient. Post CECT images were reported to compare the position of the high counts region in the SPECT images and the necrosis occurred at three month after the SIRT procedure.

### 2.7 Data analysis and representation

For all the following analyses related to the phantom study, the absorbed dose distribution obtained from the static acquisition was used as reference. The DVHs of $Ds$ and nine $Dm$ and $Dc$ configurations for both phantom and patients studies were compared using R v 4.0.2 (R Core Team, Vienna). For the patients study, the Bland-Altman methodology was used to compare  $D¯m$ and $D¯c$ of various VOIs, where the mean difference between the two measurements represented the bias, while the 95% limits of agreement (LOA), obtained by the bias ± the standard deviation multiplied by 1.96, represented the precision. Specifically for the lungs, a Bland-Altman analysis was reported to compare the $D¯c$ of the lungs and $D¯m$ lungs-2 cm, which has been proposed as a possible surrogate of the true mean dose of the lungs [
• Allred J.D.
• Niedbala J.
• Mikell J.K.
• Owen D.
• Frey K.A.
• Dewaraja Y.K.
The value of (99m)Tc-MAA SPECT/CT for lung shunt estimation in (90)Y radioembolization: a phantom and patient study.
]. Plots were obtained using R v 4.0.2 with ggplot2 library.

## 3. Results

### 3.1 Tool validation on phantom (Phantom study results)

#### 3.1.1 Visual assessment

Fig. 4 shows SPECT/CT images for the static (Table 1 - Acquisition #1, i.e., SPECTs/CT reported in panel a)) and two combinations of amplitudes and cycle durations (i.e., Acquisition #4 and #6 of Table 1) using the moving phantom (i.e., SPECTm/CT reported in panels b) and d), and SPECTc/CT shown in panels c) and e)).
A marked artifact is appreciable on both SPECTm/CT during the moving setup acquisition, leading to an elliptical shape of the high counts volume. The motion artifact is largely reduced for the SPECTc/CT image, that appears almost as a spherical volume in its higher counts region, as expected from the reference static acquisition.

#### 3.1.2 $γ$-index analysis

The $γ$-passing rate of $Dc$ was higher than $Dm$ for all the combinations of considered VOIs, amplitudes and cycle durations. Regarding the overall volume considered in the $γ$-analysis (i.e., lungs + liver), optimal (>95%) agreements were obtained for $Dc$ for the normal breathing types (5 s cycle duration and amplitudes < 20 mm), while $γ$-passing rates were acceptable (91.4%) and insufficient (87.3%) for the deep breathing amplitudes of 20 and 25 mm, respectively. Similarly, optimal $γ$-passing rates where obtained for $Dc$ for all cycle durations when an amplitude of 12.5 mm was selected with the exception of 3 s and 4 s, which nevertheless resulted in an acceptable (>90%) agreement. On the other side, $Dm$ resulted in insufficient (<90%) $γ$-passing rates for all breathing amplitudes and cycle durations when considering the overall volume (mean $γ$-passing rate: 80%).
Similar trends were observed when the liver and the lungs VOIs were analyzed separately. $Dc$ resulted in optimal agreement with $Ds$ in the liver VOI in the 5 s cycle duration for amplitudes of 5 mm, 7.5 mm and 12.5 mm and for the 40 s cycle duration for amplitude of 12.5 mm, as well as for all the cycle durations and amplitudes in the lungs VOI, with the exception of the 5 s/25 mm combination. For all the others combinations an acceptable (>90%) agreement was reached, except for the 5 s/25 mm combination that resulted in an agreement of 86% in the liver VOI. Again, an insufficient agreement between $Dm$ and $Ds$ (<90%) was obtained for all the amplitudes and cycle durations also when the two VOIs were analyzed separately, with the exception of the 5 s/5 mm combination when only the lungs VOI was considered (90%). In particular, in the 5 s/25 mm combination, the $Dm$ $γ$-passing rate in the lungs VOI was only 67%, while for $Dc$ a passing rate of 92% was obtained.
The mean $Dm$ $γ$-passing rates in the liver and in the lungs VOIs were 81% for both of them. $γ$-passing rates of $Dc$ and $Dm$ obtained on the overall (liver + lungs) VOI and on the separate liver and lungs VOIs are showed in Fig. 5.

#### 3.1.3 Isodose volumes

The absorbed dose level identyfing an isodose volume in the $Ds$ equal to the physical inner volume of the sphere (i.e., 65.5 ml) was 210 Gy. When the same dose level was applied, the 210 Gy isodose volume always differed by<10% from the physical volume of the sphere in the $Dc$ absorbed dose distribution, while differences up to 30% were observed in the $Dm$ absorbed dose distribution. The 210 Gy isodose volume of the $Dm$ was always larger than the one identified on the $Dc$, with the exception of the 5 s/5 mm and the 40 s/12.5 mm combinations, where the differences were negligible. The relative volume of the 210 Gy isodose in the $Dc$ and $Dm$ absorbed dose distribution compared to the reference dose distribution $Ds$ are reported in Fig. 6.

#### 3.1.4 Mean doses and DVHs analysis

The mean absorbed dose to the liver, tumor, lungs and lungs-2 cm VOIs obtained from the static acquisition absorbed dose distribution ($D¯s$) is reported in Table 2. For the moving ($Dm$) and corrected ($Dc$) absorbed dose distributions, the median [range] values of the mean absorbed doses across the nine amplitudes and cycle durations configurations (i.e., median  $D¯m$ and median $D¯c$) for the same VOIs are reported, together with the percentage differences from the mean doses derived from the static acquisition ($D¯s$).
Table 2SPECTs-based mean absorbed doses ($D¯s$) and median [range] of SPECTm- and SPECTc-based mean absorbed doses (i.e., $D¯m$ and $D¯c$, respectively). Median [range] percentage difference between $D¯s$ and $D¯m$ (i.e., $ΔD¯m$) and between $D¯s$ and $D¯c$ (i.e., $ΔD¯c$). The results are reported for tumor, liver, lung and lung-2 cm from the phantom study.
Median [range] of the mean doses $D¯$Median [range] percentage difference of the mean doses from the static case
$D¯s$(Gy)$D¯m$(Gy)$D¯c$(Gy)$ΔD¯m$(%)$ΔD¯c$(%)
Acquisition #1210210210210
Tumor590.7

312.6

[239.0, 338.4]
545.0

[436.3, 601.2]
−47.1

[-59.5, −42.7]
−7.7

[-26.1, 1.8]
Liver33.1

28.6

[25.6, 30.3]
33.0

[31.8, 33.4]
−13.8

[−22.7, −8.5]
−0.3

[-4.1, 0.8]
Lung1.0

19.5

[12.9, 30.4]
1.9

[1.0, 6.6]
1823.3

[1174.1, 2904.0]
86.7

[-2.9, 548.7]
Lung-2 cm2.3

[1.7, 10.4]
123.5

[-63.2, 923.2]
$D¯s$ and the median $D¯m$ and $D¯c$ to the liver were 40.2, 32.4, and 39.8 Gy respectively, while $D¯s$ and the median $D¯m$ and $D¯c$ to the tumor were 591, 338, and 549 Gy. Moreover, $D¯s$ and the median $D¯m$ and $D¯c$ to the lungs were 1.1, 18.2 and 1.6 Gy, respectively, while the median $D¯m$ to the lungs-2 cm based on SPECTm was 2.4 Gy.
The proposed tool allowed reducing the median percentage difference with respect to the reference mean dose for the liver from −19.3% ($ΔD¯m$) to −0.9% ($ΔD¯c$), while for the tumor from −42.8% ($ΔD¯m$) to −7.0% ($ΔD¯c$). The median percentage difference $ΔD¯m$ and $ΔD¯c$ for the lungs was 1586% and 47.2%, respectively. Finally, the median difference $ΔD¯m$ of the lungs-2 cm absorbed mean dose with respect to the true lungs VOI mean dose was 125%.
The use of the tool for the normal breathing type provided a tumor median $D¯c$ [range] equal to 553.1 [544.3, 601.2] Gy while the median $D¯m$ [range] was 333.8 [327.9, 339.4] Gy; similarly for the abnormal-deep breathing type the median $D¯c$ was 518.5 [436.3, 553.4] Gy while the median $D¯m$ [range] was 340.9 [264.7, 347.6] Gy. The $D¯s$ for the tumor was 590.8 Gy as reported in Table 2.
In Fig. 7 the tumor DVHs for all the amplitudes and cycle durations configurations are reported and compared with the reference DVH of the static acquisition. For all the configurations, the application of the projection images correction (i.e., the SPECTc-derived DVHs) helped to remarkably recover the reference DVH.
More in detail, Fig. 7 panels a) and b) show DVHs derived from SPECTm (left) and SPECTc (right) reconstructions both fixing the cycle duration while varying the amplitude and fixing the amplitude while varying the cycle duration, respectively. More information is given in the caption of the Figure.
Fig. 8 shows the DVH for the lungs-2 cm VOI in the SPECTm- and SPECTc-derived absorbed dose distributions and for the lungs VOI in the SPECTs-derived dose distribution, both fixing the cycle duration while varying the amplitude and fixing the amplitude while varying the cycle duration. The true lungs DVH was systematically overestimated; nevertheless, the SPECTc-derived lungs VOI DVHs better approximated the true lungs DVH with respect to the the SPECTm-derived DVHs of the lungs-2 cm VOI.

### 3.2 Tool application in a patient cohort (Patients study results)

The MATLAB tool was applied to a cohort of consecutive patients subjected to 90Y SIRT treatment. Table 3 reports the main patient and tumor characteristics including tumor type, liver and target volume, and administered activity.
Table 3Baseline characteristics of HCC patients who received 90Y TARE and whose SPECT images were reconstructed with and without the tool-based correction.
CharacteristicsPatients
Age (years) [range]71 [51, 85]
Sex (male/female)6/6
Tumor type12 HCC
Whole liver volume (ml) [range]1606 [942, 2038]
Tumor volume (ml) [range]200 [27, 334]
Tumor/whole liver ratio (%)12 [3, 21]
Distance between the hepatic dome and the upper limit of the Tumor VOI (mm) [range]17 [4, 120]
Administered activity (GBq) [range]1.0 [0.6, 2.5]
Fig. 9 illustrates the dose distribution for an example patient based on SPECT image registered with its CT image with and without the MATLAB tool-based correction. The example patient had a tumor movement along the $z$ direction with a maximum displacement of 8.8 mm from the barycentre and a distance of about 18 mm from the upper point of the hepatic dome.
From their comparison with CECT at 3-months follow-up, it is possible to observe that the 99mTc-MAA activity distribution, which was initially located near the hepatic dome, leads to an overestimation of counts in the lungs. The tool-based correction repositioned most of the activity injected in the tumor inside the liver. The CECT at 3-months follow-up shows a necrotic zone corresponding to the area in which the tool-based correction repositioned the tumor (high activity concentration areas).
Fig. 10 panels a), b), c) show the Bland-Altman analysis between $D¯m$ (SPECTm-derived mean absorbed doses) and $D¯c$ (SPECTc-derived mean absorbed doses) of tumor, liver and lungs from the patient cohort. The bias and the 95% LOA were −7.2 Gy and [-66.1 to 51.6 Gy], 0.3 Gy and [-3.4 to 4.0 Gy], and −0.3 Gy and [-4.3 to 3.6 Gy] for tumor, liver, and lungs respectively.
Fig. 10 panel d) shows the Bland-Altman analysis between $D¯m$ of lungs-2 cm and $D¯c$ of lungs from the patient cohort. The corresponding bias and the 95% LOA were 0.4 Gy and [-0.9 to 1.6 Gy].
In the patient cohort, the median DVH of SPECTm-derived absorbed dose distribution is higher than the median DVH of SPECTc-derived absorbed dose distribution in tumor VOI, especially in the high-dose region (Fig. 11). This result is in agreement with the result from the phantom study.

## 4. Discussion

In this paper, the problem of the SPECT projection images realignment and their relative shifting to the µMap has been investigated and a new calculation tool has been presented.
99mTc-MAA SPECT/CT imaging is a valuable modality for SIRT planning, although it is characterized by some limitations. To collect enough counting statistics and generate good quality images, SPECT systems are designed to minimize the distance between the patient and the detector heads. This often excludes the adoption of hardware tools, such as optical surface-guided systems. SPECT acquisitions are typically acquired in free-breathing condition across several minutes and breathing cycles. The possible drawback derived from the free-breathing condition in the SIRT pre-treatment SPECT acquisition is twofold. On one side, the SPECT system can not be prevented from being affected by the tumor motion due to the patient respiratory motion, which mainly occurs in the z (i.e., CC) direction [
• Sanders J.C.
• Ritt P.
• Kuwert T.
• Vija A.H.
• Maier A.K.
Fully Automated Data-Driven Respiratory Signal Extraction From SPECT Images Using Laplacian Eigenmaps.
]. On the other hand, the CT examinations of SPECT/CT system usually lasts few seconds, so the CT image can be considered a static-like acquisition which is generated in a random phase of the respiratory cycle. This means that the CT-derived µMap, which is used for the attenuation correction, potentially may not be aligned with the SPECT images, which are, on the contrary, averaged over all the phases of the patient respiratory cycle. This has a specific impact in liver SIRT, where the attenuation correction can strongly differ between the lungs and the liver tissues.
The proposed tool addresses both limitations. To overcome the effect of the respiratory motion, the free-breathing acquired projection images are realigned, based on the signal barycentre. Then, the signal originated in the tumor is identified on the realigned projection images by applying a threshold and a shift is performed to obtain a fine registration between the projection images and the µMap. This is automatically done by restoring the distance between the upper point of the tumor (i.e., the thresholded area on the closest projection image) and the hepatic dome, automatically identified on the µMap. In this setting, the “true” distance is defined by a previous CECT, which represents the optimal modality for target identification, and is supposed to be independent on the respiratory phase, as observed on 4D-CT [
• Paolani G.
• Strolin S.
• Santoro M.
• Della Gala G.
• Tolento G.
• Guido A.
• et al.
A novel tool for assessing the correlation of internal/external markers during SGRT guided stereotactic ablative radiotherapy treatments.
].
To validate the dosimetric impact of the tool we modified the CIRS dynamic phantom adding 3D printed and scattering components and testing several dynamic setups. To the best of our knowledge this is the first simple ad-hoc CIRS dynamic phantom adaptation created to investigate a tumor located in the hepatic dome.
The efficacy of the tool was evaluated by performing the γ-analysis with 5%/4.42 mm criteria. The γ-passing rates between the static and the corrected absorbed dose distributions in all the considered VOIs were higher than 90%, except for one amplitude and cycle duration combination, and mostly higher than 95%. On the contrary, the γ-passing rates between the static and the uncorrected dose distributions always resulted suboptimal (being about 80%).
Similary the volumes identified by the 210 Gy isodose level from the dose distributions $Dc$ and $Dm$, that in the $Ds$ corresponded to the sphere physical inner volume, were extracted. The 210 Gy isodose volumes from the dose distributions $Dc$ and $Dm$ were up to 10% and 30% of the sphere physical inner volume, respectively. Of note, the 210 Gy isodose volume of the $Dm$ dose distribution obtained with 40 s cycle duration and 12.5 mm amplitude was similar to the one of the physical sphere, but the shape and position of this isodose were different, as confirmed by the low value of the γ-passing rate. The differences in tumor mean absorbed doses between the reference dose distribution and the tool-corrected (i.e., $ΔD¯c$) and uncorrected (i.e., $ΔD¯m$) dose distributions were<10% and up to 43%, respectively, while for the liver were<1% and up to 19%, respectively. Thus, the correction provided a more accurate absorbed dose estimation including all the investigated amplitudes and cycle durations. For both the subset of amplitudes and cycles representative of normal or abnormal-deep breathing types, the DVHs derived from the corrected images were closer to the ones obtained from the static acquisition.
The largest percentage impact of the corrected images was observed on lungs and lungs-2 cm VOI dosimetry. Indeed, for the lungs we observed a mean dose overestimation reduction from $ΔD¯m$  = 1586% to $ΔD¯c$ = 47%. Of note, using the lungs-2 cm VOI, the $ΔD¯m$ was 125%.
It might be pointed out that the lungs-like VOI in the phantom was smaller than the lungs of a typical patient (being about 1080 ml vs. a typical volume of 3500 ml), but the relative mean doses remain representative of the investigated effect.
The tool was also prospectively applied on a cohort of patients undergoing 99mTc-MAA SPECT/CT before 90Y resin microsphere radioembolization. CECT images collected during the patients' follow-up were analyzed to investigate the potential impact of the proposed approach on our clinical practice. Post-treatment 90Y PET/CT are also available and will be used for a separate paper.
Regarding the patient cohort, variations (+/- 50 Gy) and (+/- 4 Gy) were observed for SPECTc- and SPECTm-derived mean absorbed doses of tumor and liver, respectively. These results show that tumor mean doses had a wider range of variability than the liver, since most of the activity, although mispositioned, still was inside this VOI.
SPECTc- derived mean absorbed doses of tumor mean doses were lower than SPECTm- derived ones, while the contrary was observed for the liver. This behaviour agrees with our phantom study for the liver, but not for the tumor. This is likely due to the large variability of the distance of the tumor barycentre from the hepatic dome position, as reported in Table 3. In other words, in our phantom study the tumor was very close (<1 cm) to the hepatic dome-like position. On the other hand, a possible explanation of the agreement between patient and phantom study for the liver is that the correction performed with the MATLAB tool relocated counts, originally positioned in the lungs, inside the liver, increasing its mean dose.
Lungs mean absorbed doses showed a small range of variability. SPECTc-derived mean absorbed doses of the lungs were lower than SPECTm-derived ones, while they were higher than the SPECTm-derived mean absorbed doses of lungs-2 cm. This tendency is the opposite of what was observed in the phantom study, although the mean dose underestimation in the lungs-2 cm VOI can potentially lead to an under-estimation of the risk of lungs toxicity.
DVHs of phantom-based corrected images showed an increase in high dose tumoral region in the phantom study and in the illustrative patient cohort.
Clearly, when comparing the SPECTc and SPECTm-derived absorbed dose distributions on patients, the true dose distribution obtained in the static condition is missing. Thus, considering the passive targeting of 90Y microsphere treatment, which are attached to the tumor target independently of the tumor motion, the radiological response based on CECT follow-up images should be investigated. Images from an example patient show the agreement between the corrected SPECT/CT-derived dosimetry and the marked areas of treatment response. More patients are needed to confirm this result. A more general validation of the impact of the tool on the dose distribution in larger patients cohort and on the subsequent dose–effect relationship (correlation between tumor dose and tumor response) will be addressed in future studies.
The tool presents several advantages. First, only the projection images belonging to the detector head closer to the tumor is considered for the realignment, allowing the correction not to be influenced by the other detector head, in which the scatter contribution is much higher and the projection image barycentre $Btpj,k$ is more difficult to detect. The selection of the most proper detector head is taken automatically by the MATLAB tool at the point c) based on geometric considerations.
In addition, our tool is vendor-independent and potentially useful for several SPECT/CT systems allowing exporting and re-importing the corrected raw data for image reconstruction. The tool requires as input only the projection images of the SPECT acquisition, the tumor and liver contours in DICOM RT-struct format, the CT and the derived µMap. All of these data are typically easily accessible in SPECT/CT system. The use of the µMap instead of the CT in the MATLAB app is justified by the fact that the µMap is already sampled with the same pixel size of the projection images, but the tool can be easily adapted to resample the CT when the µMap is not available.
After the correction, our tool allows saving the realigned/shifted projection images in DICOM format, which can also be used as input of reconstruction system of SPECT/CT vendor (i.e., GE Helthcare) or an independent image reconstruction systems, such as MIM software (data not shown).
A possible limitation of our study is that the phantom study has a restricted realism as compared with the complexity of clinical studies. The results of the tool validation on patients might be affected by considerable uncertainties due to the large variability in tumor, liver or lungs volumes. The efficacy of this method might be limited when the activity distribution is heterogeneous inside the tumor, or large necrotic areas are present within the target lesion. For the same reason, it has been currently applied only in patients with a not hypovascular/infiltrative or disomogeneous HCC lesion.
It is worth noting that patients with high LSF (>20%) evaluated with the planar scintigraphy approach are excluded from the treatment and no patients with high LSF are included in our cohort. Further evaluation might be needed for these infrequent cases as reported in another study [
• Bastiaannet R.
• van der Velden S.
• Lam M.G.E.H.
• Viergever M.A.
• de Jong H.W.A.M.
Fast and accurate quantitative determination of the lung shunt fraction in hepatic radioembolization.
]. Possible improvements to our clinical workflow can be suggested both in the execution of the exam and in the image correction and reconstruction. The acquisition of quasi-static CT images at the mid-phase (i.e., between the minimum inspiratory phase and the maximum expiratory phase) could be an option to be implemented. Unfortunately, the patient should also hold the breathing in the mid-phase, but this is difficult to realize and would require an active breathing controller which is not currently available in most SPECT/CT devices. The use of helical CT acquired at end-expiration has been suggested for LSF estimation [
• Lu Z.
• Chen G.
• Lin K.-H.
• Wu T.-H.
• Mok G.S.P.
Evaluation of different CT maps for attenuation correction and segmentation in static (99m) Tc-MAA SPECT/CT for (90) Y radioembolization treatment planning: A simulation study.
], although SPECT/CT alignment mismatch may arise with the risk of overestimating the LSF and excluding potential candidates to the treatment.
In addition, fast 99mTc-MAA SPECT/CT protocol might also be implemented and tested with our tool/moving phantom. The feasibility of this approach relies on the observation that SNR does not affect SPECT/CT accuracy for the LSF evaluation, as reported in [
• van der Velden S.
• Dietze M.M.A.
• Viergever M.A.
• de Jong H.W.A.M.
Fast technetium-99m liver SPECT for evaluation of the pretreatment procedure for radioembolization dosimetry.
]. SPECT reconstructions are also affected by partial volume effects, causing the activity spill-in and spill-out phenomena in lungs/liver or tumor, resulting in an over-and under-activity estimation, respectively [
• Lu Z.
• Chen G.
• Lin K.-H.
• Wu T.-H.
• Mok G.S.P.
Evaluation of different CT maps for attenuation correction and segmentation in static (99m) Tc-MAA SPECT/CT for (90) Y radioembolization treatment planning: A simulation study.
]. This might be considered in the absorbed dose distribution evaluation to improve the accuracy of the tumor absorbed dose estimation.
Furhermore, our tool allows the reduction of the impact due to the respiratory movements by the tracking of the target and/or surrogates (e.g., such as the hepatic dome [
• Paolani G.
• Strolin S.
• Santoro M.
• Della Gala G.
• Tolento G.
• Guido A.
• et al.
A novel tool for assessing the correlation of internal/external markers during SGRT guided stereotactic ablative radiotherapy treatments.
]) without using external respiratory tracking systems, which are critical to be implemented in SPECT technology due to the closeness of the detectors to the patient to maximize counts/resolution.
Finally, our tool might be included in the attempts of decreasing the uncertanties of planning in SIRT treatments through an accurate image quantification. The image analysis performed by using the γ-index and isodoses approaches allows a better explanation of mean absorbed doses and DVHs which, as seen in external beam radiotherapy, are related to patient clinical outcomes [
• Garin E.
• Tselikas L.
• Guiu B.
• Chalaye J.
• Edeline J.
• de Baere T.
• et al.
Personalised versus standard dosimetry approach of selective internal radiation therapy in patients with locally advanced hepatocellular carcinoma (DOSISPHERE-01): a randomised, multicentre, open-label phase 2 trial.
,
• Strigari L.
• Konijnenberg M.
• Chiesa C.
• Bardies M.
• Du Y.
• Gleisner K.S.
• et al.
The evidence base for the use of internal dosimetry in the clinical practice of molecular radiotherapy.
].

## 5. Conclusion

For radioembolization, the liver lesion is passively targeted by the activity injection inside the area, while the activity quantification is still an issue for balancing its efficacy and toxicity. The proposed tool allows the correction of the respiratory motion artifacts detected in 99mTc-MAA SPECT/CT images and potential improvement of the accuracy obtained in dose estimation to the target and normal tissue for a more precise determination of the activity to be injected during the SIRT procedure.

## Funding

The study was partially supported by AIRC IG 20809 (2017) PI: L.Strigari.

## Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

## Acknowledgements

The authors gratefully acknowledge Prof. Emanuela Marcelli and Eng. Barbara Bortolani from the eDIMES Lab-Laboratory of Bioengineering (Department of Experimental, Diagnostic and Specialty Medicine, University of Bologna, Italy) for the production of 3D-printed components shown in the phantom section of the manuscript.

## References

• Atassi B.
• Bangash A.K.
• Bahrani A.
• Pizzi G.
• Lewandowski R.J.
• Ryu R.K.
• et al.
Multimodality Imaging Following 90Y Radioembolization: A Comprehensive Review and Pictorial Essay.
• Smits M.L.J.
• Dassen M.G.
• Prince J.F.
• Braat A.J.A.T.
• Beijst C.
• Bruijnen R.C.G.
• et al.
The superior predictive value of (166)Ho-scout compared with (99m)Tc-macroaggregated albumin prior to (166)Ho-microspheres radioembolization in patients with liver metastases.
Eur J Nucl Med Mol Imaging. 2020; 47: 798-806
• Chiesa C.
• Maccauro M.
(166)Ho microsphere scout dose for more accurate radioembolization treatment planning.
Eur J Nucl Med Mol Imaging. 2020; 47: 744-747
• Cremonesi M.
• Chiesa C.
• Strigari L.
• Ferrari M.
• Botta F.
• Guerriero F.
• et al.
Front Oncol. 2014; 4
• Elsayed M.
• Cheng B.
• Xing M.
• Sethi I.
• Brandon D.
• Schuster D.M.
• et al.
Comparison of Tc-99m MAA Planar Versus SPECT/CT Imaging for Lung Shunt Fraction Evaluation Prior to Y-90 Radioembolization: Are We Overestimating Lung Shunt Fraction?.
Cardiovasc Intervent Radiol. 2021; 44: 254-260
• Chiesa C.
• Lambert B.
• Maccauro M.
• Ezziddin S.
• Dieudonné A.
• et al.
Pretreatment Dosimetry in HCC Radioembolization with (90)Y Glass Microspheres Cannot Be Invalidated with a Bare Visual Evaluation of (99m)Tc-MAA Uptake of Colorectal Metastases Treated with Resin Microspheres.
J Nucl Med. 2014; 55: 1215-1216
• Ulrich G.
• Dudeck O.
• Furth C.
• Ruf J.
• Grosser O.S.
• et al.
Predictive value of intratumoral 99mTc-macroaggregated albumin uptake in patients with colorectal liver metastases scheduled for radioembolization with 90Y-microspheres.
J Nucl Med. 2013; 54: 516-522
• Allred J.D.
• Niedbala J.
• Mikell J.K.
• Owen D.
• Frey K.A.
• Dewaraja Y.K.
The value of (99m)Tc-MAA SPECT/CT for lung shunt estimation in (90)Y radioembolization: a phantom and patient study.
EJNMMI Res. 2018; 8: 50
• Kao Y.H.
• Magsombol B.M.
• Toh Y.
• Tay K.H.
• Chow P.KH.
• Goh A.SW.
• et al.
Personalized predictive lung dosimetry by technetium-99m macroaggregated albumin SPECT/CT for yttrium-90 radioembolization.
EJNMMI Res. 2014; 4
• Yu N.
• Srinivas S.M.
• Difilippo F.P.
• Shrikanthan S.
• Levitin A.
• McLennan G.
• et al.
Lung dose calculation with SPECT/CT for 90Yittrium radioembolization of liver cancer.
Int J Radiat Oncol Biol Phys. 2013; 85: 834-839
• Chiesa C.
• Sjogreen-Gleisner K.
• Walrand S.
• Strigari L.
• Flux G.
• Gear J.
• et al.
EANM dosimetry committee series on standard operational procedures: a unified methodology for (99m)Tc-MAA pre- and (90)Y peri-therapy dosimetry in liver radioembolization with (90)Y microspheres.
EJNMMI Phys. 2021; 8
• Dewaraja Y.K.
• Frey E.C.
• Sgouros G.
• Brill A.B.
• Roberson P.
• Zanzonico P.B.
• et al.
MIRD pamphlet No. 23: quantitative SPECT for patient-specific 3-dimensional dosimetry in internal radionuclide therapy.
J Nucl Med. 2012; 53: 1310-1325
• Chiesa C.
• Strigari L.
• Pacilio M.
• Richetta E.
• Cannatà V.
• Stasi M.
• et al.
Dosimetric optimization of nuclear medicine therapy based on the Council Directive 2013/59/EURATOM and the Italian law N. 101/2020. Position paper and recommendations by the Italian National Associations of Medical Physics (AIFM) and Nuclear Medicine (AIMN).
Physica Med. 2021; 89: 317-326
• Walrand S.
• Hesse M.
• Amini N.
• Lhommel R.
• et al.
Accurate non-tumoral 99mTc-MAA absorbed dose prediction to plan optimized activities in liver radioembolization using resin microspheres.
Phys Med. 2021; 89: 250-257
• Garin E.
• Tselikas L.
• Guiu B.
• Chalaye J.
• Edeline J.
• de Baere T.
• et al.
Personalised versus standard dosimetry approach of selective internal radiation therapy in patients with locally advanced hepatocellular carcinoma (DOSISPHERE-01): a randomised, multicentre, open-label phase 2 trial.
Lancet Gastroenterol Hepatol. 2021; 6: 17-29
• Strigari L.
• Sciuto R.
• Rea S.
• Carpanese L.
• Pizzi G.
• Soriani A.
• et al.
Efficacy and toxicity related to treatment of hepatocellular carcinoma with 90Y-SIR spheres: radiobiologic considerations.
J Nucl Med. 2010; 51: 1377-1385
• Bernard C.
• Lovinfosse P.
• Gérard L.
• Lilet H.
• Cornet O.
• et al.
Comparative dosimetry between (99m)Tc-MAA SPECT/CT and (90)Y PET/CT in primary and metastatic liver tumors.
Eur J Nucl Med Mol Imaging. 2020; 47: 828-837
• Kafrouni M.
• Allimant C.
• Vauclin S.
• Guiu B.
• Mariano-Goulart D.
• et al.
Analysis of differences between (99m)Tc-MAA SPECT- and (90)Y-microsphere PET-based dosimetry for hepatocellular carcinoma selective internal radiation therapy.
EJNMMI Res. 2019; 9
• Kim S.P.
• Cohalan C.
• Kopek N.
• Enger S.A.
A guide to (90)Y radioembolization and its dosimetry.
Phys Med. 2019; 68: 132-145
• Bastiaannet R.
• Viergever M.A.
• de Jong H.W.A.M.
Impact of respiratory motion and acquisition settings on SPECT liver dosimetry for radioembolization.
Med Phys. 2017; 44: 5270-5279
• Bitarafan-Rajabi A.
• Rajabi H.
• Rastgou F.
• Yaghoobi N.
• Malek H.
• et al.
Influence of respiratory motion correction on quantification of myocardial perfusion SPECT.
J Nucl Cardiol. 2015; 22: 1019-1030
• Dasari P.K.R.
• Könik A.
• Pretorius P.H.
• Johnson K.L.
• Segars W.P.
• Shazeeb M.S.
• et al.
Correction of hysteretic respiratory motion in SPECT myocardial perfusion imaging: Simulation and patient studies.
Med Phys. 2017; 44: 437-450
• Daou D.
• Sabbah R.
• Coaguila C.
• Boulahdour H.
Feasibility of data-driven cardiac respiratory motion correction of myocardial perfusion CZT SPECT: A pilot study.
J Nucl Cardiol. 2017; 24: 1598-1607
• Kovalski G.
• Israel O.
• Keidar Z.
• Frenkel A.
• Sachs J.
• Azhari H.
Correction of heart motion due to respiration in clinical myocardial perfusion SPECT scans using respiratory gating.
J Nucl Med. 2007; 48: 630-636
• Polycarpou I.
• Chrysanthou-Baustert I.
• Parpottas Y.
• Panagidis C.
• Marsden P.K.
• et al.
Impact of respiratory motion correction on SPECT myocardial perfusion imaging using a mechanically moving phantom assembly with variable cardiac defects.
J Nucl Cardiol. 2017; 24: 1216-1225
• Segars W.P.
• Mok S.P.
• Tsui B.M.W.
Investigation of Respiratory Gating in Quantitative Myocardial SPECT.
IEEE Trans Nucl Sci. 2009; 56: 91-96
1. Healthcare G. Motion Detection and Correction (MDC) on Xeleris: white paper.

• Sanders J.C.
• Ritt P.
• Kuwert T.
• Vija A.H.
• Maier A.K.
Fully Automated Data-Driven Respiratory Signal Extraction From SPECT Images Using Laplacian Eigenmaps.
IEEE Trans Med Imaging. 2016; 35: 2425-2435
• Robert A.
• Rit S.
• Baudier T.
• Jomier J.
• Sarrut D.
Data-Driven Respiration-Gated SPECT for Liver Radioembolization.
IEEE Transactions on Radiation and Plasma Medical Sciences. 2022;
• Ritt P.
Recent Developments in SPECT/CT.
Semin Nucl Med. 2022; 52: 276-285
• van der Vos C.S.
• Koopman D.
• Rijnsdorp S.
• Arends A.J.
• Boellaard R.
• van Dalen J.A.
• et al.
Quantification, improvement, and harmonization of small lesion detection with state-of-the-art PET.
Eur J Nucl Med Mol Imaging. 2017; 44: 4-16
• Dietze M.M.A.
• Bastiaannet R.
• Kunnen B.
• Velden S.
• Lam M.G.E.H.
• Viergever M.A.
• et al.
Respiratory motion compensation in interventional liver SPECT using simultaneous fluoroscopic and nuclear imaging.
Med Phys. 2019; 46: 3496-3507
• Dietze M.M.A.
• Kunnen B.
• Brontsema F.
• Ramaekers P.
• Beijst C.
• Afifah M.
• et al.
A compact and mobile hybrid C-arm scanner for simultaneous nuclear and fluoroscopic image guidance.
• Dietze M.M.A.
• Kunnen B.
• Lam M.G.E.H.
• de Jong H.W.A.M.
Interventional respiratory motion compensation by simultaneous fluoroscopic and nuclear imaging: a phantom study.
Phys Med Biol. 2021; 66: 065001
• Keall P.J.
• Mageras G.S.
• Balter J.M.
• Emery R.S.
• Forster K.M.
• Jiang S.B.
• et al.
The management of respiratory motion in radiation oncology report of AAPM Task Group 76a).
Med Phys. 2006; 33: 3874-3900
• Yuan G.
• Drost N.A.
• McIvor R.A.
Respiratory rate and breathing pattern.
McMaster Univ Med J. 2013; 10: 23-28
• Low D.A.
• Harms W.B.
• Mutic S.
• Purdy J.A.
A technique for the quantitative evaluation of dose distributions.
Med Phys. 1998; 25: 656-661
• Hussein M.
• Clark C.H.
• Nisbet A.
Challenges in calculation of the gamma index in radiotherapy - Towards good practice.
Phys Med. 2017; 36: 1-11
• Miften M.
• Olch A.
• Mihailidis D.
• Moran J.
• Pawlicki T.
• Molineu A.
• et al.
Tolerance limits and methodologies for IMRT measurement-based verification QA: Recommendations of AAPM Task Group No. 218.
Med Phys. 2018; 45: e53-e83
• Maughan N.M.
• Garcia‐Ramirez J.
• Arpidone M.
• Swallen A.
• Laforest R.
• Goddu S.M.
• et al.
Validation of post-treatment PET-based dosimetry software for hepatic radioembolization of Yttrium-90 microspheres.
Med Phys. 2019; 46: 2394-2402
• Paolani G.
• Strolin S.
• Santoro M.
• Della Gala G.
• Tolento G.
• Guido A.
• et al.
A novel tool for assessing the correlation of internal/external markers during SGRT guided stereotactic ablative radiotherapy treatments.
Phys Med. 2021; 92: 40-51
• Bastiaannet R.
• van der Velden S.
• Lam M.G.E.H.
• Viergever M.A.
• de Jong H.W.A.M.
Fast and accurate quantitative determination of the lung shunt fraction in hepatic radioembolization.
Phys Med Biol. 2019; 64: 235002
• Lu Z.
• Chen G.
• Lin K.-H.
• Wu T.-H.
• Mok G.S.P.
Evaluation of different CT maps for attenuation correction and segmentation in static (99m) Tc-MAA SPECT/CT for (90) Y radioembolization treatment planning: A simulation study.
Med Phys. 2021; 48: 3842-3851
• van der Velden S.
• Dietze M.M.A.
• Viergever M.A.
• de Jong H.W.A.M.
Fast technetium-99m liver SPECT for evaluation of the pretreatment procedure for radioembolization dosimetry.
Med Phys. 2019; 46: 345-355
• Strigari L.
• Konijnenberg M.
• Chiesa C.
• Bardies M.
• Du Y.
• Gleisner K.S.
• et al.
The evidence base for the use of internal dosimetry in the clinical practice of molecular radiotherapy.
Eur J Nucl Med Mol Imaging. 2014; 41: 1976-1988