Simulation CT Scan Research Articles

Technical note: Phantom-based evaluation of CBCT dose calculation accuracy for use in adaptive radiotherapy.

High-quality 3D-anatomy of the day is needed for treatment plan adaptation in radiotherapy. For online x-ray-based CBCT workflows, one approach is to create a synthetic CT or to utilize a fan-beam CT with corresponding registrations. The former potentially introduces uncertainties in the dose calculation if deformable image registration is used. The latter can introduce burden and complexity to the process, the facility, and thepatient. Using the CBCT of the day, acquired on the treatment device, for direct dose calculation and plan adaptation can overcome these limitations. This study aims to assess the accuracy of the calculated dose on the CBCT scans acquired on a Halcyon linear accelerator equipped withHyperSight. HyperSight's new CBCT reconstruction algorithm includes improvements in scatter correction, HU calibration of the imager, and beam shape adaptation. Furthermore, HyperSight introduced a new x-ray detector. To show the effect of the implemented improvements, gamma comparisons of 2%/2mm, 2%/1mm, and 1%/1mm were made between the dose distribution in phantoms calculated on the CBCT reconstructions and the simulation CT scans, considering this the standard of care. The resulting gamma passing rates were compared to those obtained with the Halcyon 3.0 reconstruction and hardware without HyperSight's technologies. Various anatomical phantoms for dosimetric evaluations on brain, head and neck, lung, breast, and prostate cases have been used in thisstudy. The overall results demonstrated that HyperSight outperformed the Halcyon 3.0 version. Based on the gamma analysis, the calculated dose using HyperSight was closer to the CT scan-based doses than the calculated dose using iCBCT Halcyon 3.0 for most cases. Over all plans and gamma criteria, Halcyon 3.0 achieved an average passing rate of 92.9%, whereas HyperSight achieved 98.1%. Using HyperSight CBCT images for direct dose calculation, for example, in (online) plan adaptation, seems feasible for the investigatedcases.

Impact of Simulated Reduced-Dose Chest CT on Diagnosing Pulmonary T1 Tumors and Patient Management.

To determine the diagnostic performance of simulated reduced-dose chest CT scans regarding pulmonary T1 tumors and assess the potential impact on patient management, a repository of 218 patients with histologically proven pulmonary T1 tumors was used. Virtual reduced-dose images were simulated at 25%- and 5%-dose levels. Tumor size, attenuation, and localization were scored by two experienced chest radiologists. The impact on patient management was assessed by comparing hypothetical LungRADS scores. The study included 210 patients (41% females, mean age 64.5 ± 9.2 years) with 250 eligible T1 tumors. There were differences between the original and the 5%-but not the 25%-dose simulations, and LungRADS scores varied between the dose levels with no clear trend. Sensitivity of Reader 1 was significantly lower using the 5%-dose vs. 25%-dose vs. original dose for size categorization (0.80 vs. 0.85 vs. 0.84; p = 0.007) and segmental localization (0.81 vs. 0.86 vs. 0.83; p = 0.018). Sensitivities of Reader 2 were unaffected by a dose reduction. A CT dose reduction may affect the correct categorization and localization of pulmonary T1 tumors and potentially affect patient management.

Improving Automated Hemorrhage Detection at Sparse-View CT via U-Net-based Artifact Reduction.

Purpose To explore the potential benefits of deep learning-based artifact reduction in sparse-view cranial CT scans and its impact on automated hemorrhage detection. Materials and Methods In this retrospective study, a U-Net was trained for artifact reduction on simulated sparse-view cranial CT scans in 3000 patients, obtained from a public dataset and reconstructed with varying sparse-view levels. Additionally, EfficientNet-B2 was trained on full-view CT data from 17 545 patients for automated hemorrhage detection. Detection performance was evaluated using the area under the receiver operating characteristic curve (AUC), with differences assessed using the DeLong test, along with confusion matrices. A total variation (TV) postprocessing approach, commonly applied to sparse-view CT, served as the basis for comparison. A Bonferroni-corrected significance level of .001/6 = .00017 was used to accommodate for multiple hypotheses testing. Results Images with U-Net postprocessing were better than unprocessed and TV-processed images with respect to image quality and automated hemorrhage detection. With U-Net postprocessing, the number of views could be reduced from 4096 (AUC: 0.97 [95% CI: 0.97, 0.98]) to 512 (0.97 [95% CI: 0.97, 0.98], P < .00017) and to 256 views (0.97 [95% CI: 0.96, 0.97], P < .00017) with a minimal decrease in hemorrhage detection performance. This was accompanied by mean structural similarity index measure increases of 0.0210 (95% CI: 0.0210, 0.0211) and 0.0560 (95% CI: 0.0559, 0.0560) relative to unprocessed images. Conclusion U-Net-based artifact reduction substantially enhanced automated hemorrhage detection in sparse-view cranial CT scans. Keywords: CT, Head/Neck, Hemorrhage, Diagnosis, Supervised Learning Supplemental material is available for this article. © RSNA, 2024.

Diagnostic CT-Enabled Planning (DART): Results of a Randomized Trial in Palliative Radiation Therapy

Using diagnostic computed tomography (dCT) scans instead of CT simulation (CTsim) scans can increase departmental efficiency and reduce patient burden. The goal of the DART trial was to assess the efficacy and acceptability of dCT-based planning workflows with a focus on patient experiences, plan deliverability and adequacy of target coverage, and workflows. Patients undergoing same-day CTsim and treatment for palliative radiation therapy to thoracic, abdominopelvic, or proximal limb targets with a recent dCT (within 28 days) in a reproducible position were eligible. After stratifying by target type (bone or soft tissue vs. visceral), participants were randomized (1:2 ratio) between CTsim-based (CTsim arm) vs. dCT-based planning (dCT arm). The primary endpoint was time in center (TIC), defined as total time spent in the cancer center on first day of treatment, from first radiation department appointment to first fraction completion. Secondary endpoints included plan deliverability, adequacy of target coverage, and stakeholder acceptability. Thirty-three patients (42 treatment sites) were enrolled between June 2022 and April 2023. The median age was 72 (interquartile range [IQR]: 67-78), 73% were male, and the most common primary cancers were lung (33%), prostate (24%), and breast (12%). The most common dose and fractionations were 8 Gy in 1 and 20 Gy in 5 fractions (50% and 43% of plans, respectively). TIC was 4.7 ± 1.1 hours (mean ± SD) in the CTsim arm vs. 0.41 ± 0.14 hours in the dCT arm (P < .001). All dCT plans were deliverable. All plans in both arms were rated as "acceptable" (80% CTsim; 81% dCT) or "acceptable with minor deviation" (20% CTsim; 19% dCT). Patient perception of acceptability was similar in both arms with the exception of time burden, which was rated as "acceptable" by 50% in the CTsim arm vs. 90% in the dCT arm (P = .025). dCT-based radiation planning substantially reduced TIC without detriment in plan deliverability or quality and had a tangible impact on patient experience with reduced patient-reported time burden.

A prospective study on the effect of tumor shrinkage on exit fluence gamma pass rate in high precision radiotherapy and influence of phantom setup error in patient-specific quality assurance.

Objective parameters for decision on adaptive radiotherapy depend on patient, tumor and treatment related factors. Present study reports geometric uncertainties occurring during high precision radiotherapy, beam fluence analysis and serial exit dose measurement as a patient-specific tool for adaptive radiotherapy. Serial exit dose fluence of 24 patients (at baseline and mid-treatment) undergoing IMRT/VMAT treatment were measured. Baseline and midtreatment exit dose evaluation was done using gafchromic films in predefined region of interest. Difference of volume of GTV at baseline (from simulation CT scan) and midtreatment CBCT scan was calculated (ΔGTV). Population based systematic errors (mm) were 4.15, 2.26, 0.88 and random errors (mm) were 2.56, 3.69, and 2.03 in mediolateral (ML), craniocaudal (CC) and anteroposterior (AP) directions respectively. Gamma pass rate reduced with incremental shift. For a 5 mm shift, maximum deviation was found in anteroposterior axis (22.16 ± 7.50) and lowest in mediolateral axis (12.85 ± 4.95). On serial measurement of exit dose fluence, tumor shrinkage significantly influenced gamma pass rate. The mean gamma pass rate was significantly different between groups with 50% shrinkage of tumor volume (86.36 vs 96.24, P = 0.008, on multivariate analysis P = 0.026). Rapid fall of gamma pass rate was observed for set up error of ≥3 mm. Serial measurement of exit dose fluence by radiochromic film is a feasible method of exit dose comparison in IMRT/VMAT, where EPID dosimetry is not available with linear accelerator configuration. Our study suggests that there is a significant difference between gamma pass rates of baseline and mid treatment exit dose fluence with greater than 50% tumor shrinkage.

Dosimetric feasibility study (“proof of concept”) of refractory ventricular tachycardia radioablation using proton minibeams

PurposePreclinical data demonstrated that the use of proton minibeam radiotherapy reduces the risk of toxicity in healthy tissue. Ventricular tachycardia radioablation is an area under clinical investigation in proton beam therapy. We sought to simulate a ventricular tachycardia radioablation with proton minibeams and to demonstrate that it was possible to obtain a homogeneous coverage of an arrhythmogenic cardiac zone with this technique. Material and methodsAn arrhythmogenic target volume was defined on the simulation CT scan of a patient, localized in the lateral wall of the left ventricle. A dose of 25Gy was planned to be delivered by proton minibeam radiotherapy, simulated using a Monte Carlo code (TOPAS v.3.7) with a collimator of 19 0.4 mm-wide slits spaced 3mm apart. The main objective of the study was to obtain a plan ensuring at least 93% of the prescription dose in 93% of the planning target volume without exceeding 110% of the prescribed dose in the planning target volume. ResultsThe average dose in the planning treatment volume in proton minibeam radiotherapy was 25.12Gy. The percentage of the planning target volume receiving 93% (V93%), 110% (V110%), and 95% (V95%) of the prescribed dose was 94.25%, 0%, and 92.6% respectively. The lateral penumbra was 6.6mm. The mean value of the peak-to-valley-dose ratio in the planning target volume was 1.06. The mean heart dose was 2.54Gy versus 5.95Gy with stereotactic photon beam irradiation. ConclusionThis proof-of-concept study shows that proton minibeam radiotherapy can achieve a homogeneous coverage of an arrhythmogenic cardiac zone, reducing the dose at the normal tissues. This technique, ensuring could theoretically reduce the risk of late pulmonary and breast fibrosis, as well as cardiac toxicity as seen in previous biological studies in proton minibeam radiotherapy.

SC-GAN: Structure-completion generative adversarial network for synthetic CT generation from MR images with truncated anatomy

Creating synthetic CT (sCT) from magnetic resonance (MR) images enables MR-based treatment planning in radiation therapy. However, the MR images used for MR-guided adaptive planning are often truncated in the boundary regions due to the limited field of view and the need for sequence optimization. Consequently, the sCT generated from these truncated MR images lacks complete anatomic information, leading to dose calculation error for MR-based adaptive planning. We propose a novel structure-completion generative adversarial network (SC-GAN) to generate sCT with full anatomic details from the truncated MR images. To enable anatomy compensation, we expand input channels of the CT generator by including a body mask and introduce a truncation loss between sCT and real CT. The body mask for each patient was automatically created from the simulation CT scans and transformed to daily MR images by rigid registration as another input for our SC-GAN in addition to the MR images. The truncation loss was constructed by implementing either an auto-segmentor or an edge detector to penalize the difference in body outlines between sCT and real CT. The experimental results show that our SC-GAN achieved much improved accuracy of sCT generation in both truncated and untruncated regions compared to the original cycleGAN and conditional GAN methods.

Electromagnetic Transmitter-Based Prostate Gating for Dose-Escalated Linac-Based Stereotactic Body Radiation Therapy: An Evaluation of Intrafraction Motion.

Stereotactic Body Radiotherapy (SBRT) is as a standard treatment for prostate cancer (PCa). Tight margins and high dose gradients are needed, and the precise localization of the target is mandatory. Our retrospective study reports our experience regarding the evaluation of intrafraction prostate motion during LINAC-based SBRT evaluated with a novel electromagnetic (EM) tracking device. This device consists of an integrated Foley catheter with a transmitter connected to a receiver placed on the treatment table. We analyzed 31 patients who received LINAC-based SBRT using flattening filter-free (FFF) volumetric modulated arc therapy (VMAT). The patients were scheduled to be treated for primary (n = 27) or an intraprostatic recurrent PCa (n = 4). A simulation CT scan was conducted while the patients had a filled bladder (100-150 cc) and an empty rectum, and an EM tracking device was used. The same rectal and bladder conditions were employed during the treatment. The patients received 36.25 Gy delivered over five consecutive fractions on the whole prostate and 40 Gy on the nodule(s) visible via MRI, both delivered with a Simultaneous Integrated Boost approach. The CTV-to-PTV margin was 2 mm for both the identified treatment volumes. Patient positioning was verified with XVI ConeBeam-CT (CBCT) matching before each fraction. When the signals exceeded a 2 mm threshold in any of the three spatial directions, the treatment was manually interrupted. A new XVI CBCT was performed if this offset lasted >20 s. We analyzed data about 155 fractions. The median and mean treatment times, calculated per fraction, were 10 m31 s and 12 m44 s (range: 6 m36 s-65 m28 s), and 95% of the fractions were delivered with a maximum time of 27 m48 s. During treatment delivery, the mean and median number of XVI CBCT operations realized during the treatment were 2 and 1 (range: 0-11). During the treatment, the prostate was outside the CTV-to-PTV margin (2 mm), thus necessitating the stoppage of the delivery +/- a reacquisition of the XVI CBCT for 11.2%, 8.9%, and 3.9% of the delivery time in the vertical, longitudinal, and lateral direction, respectively. We easily integrated an EM-transmitter-based gating for prostate LINAC-based SBRT into our normal daily workflow. Using this system, a 2 mm CTV-to-PTV margin could be safely applied. A small number of fractions showed a motion exceeding the predefined 2 mm threshold, which would have otherwise gone undetected without intrafraction motion management.

Can physical parameters from radiation simulation scan with deep inspiratory breath hold predict magnitude of heart dose reduction?

ABSTRACT Introduction: Deep inspiratory breath hold is one of the techniques for reducing the heart doses for left breast cancers. This study was conducted to confirm use of physical parameters from DIBH simulation CT scan like DIBH amplitude alongside several novel parameters to predict the heart dose reduction. Materials and Methods: Segmentation and planning of radiation to the left breast on the free breathing (FB) and DIBH simulation scan was performed for 50 left-sided breast cancer patients treated with DIBH technique. Physical parameters, namely DIBH amplitude, anterior sternal displacement, diaphragmatic excursion, ratio of lung volume (cc) in DIBH scan to lung volume in FB scan (cc), and delta heart volume in field (DHVIF), were extracted and were compared with magnitude of heart dose reduction (mean heart dose, V30Gy, and D5%). Results: Forty-eight (96%) patients achieved reduction in the mean heart dose with DIBH technique, while all patients had reduction in V30Gy. The median reduction was 41%, 89.7%, and 63% in the mean dose, V30Gy, and D5%, respectively. While DIBH did not correlate with heart dose reduction, ratio of lung volumes and DHVIF showed a strong positive correlation with heart dose reduction (P &lt; 0.05). Sternal displacement correlated weakly with heart dose reduction but strongly with DHVIF, demonstrating to be an indirect predictor. Conclusions: Physical parameters like anterior sternal displacement, ratio of lung volumes of DIBH to FB, and possibly diaphragmatic movement can predict the dose reduction before the dose calculations by the physicist. These parameters can be used to construct a model to predict heart dose reduction.

A preliminary study of dynamic interactive simulation and computational CT scan of the ideal alveolus model.

While the development of CT imaging technique has brought cognition of in vivo organs, the resolution of CT images and their static characteristics have gradually become barriers of microscopic tissue research. Previous research used the finite element method to study the airflow and gas exchange in the alveolus and acinar to show the fate of inhaled aerosols and studied the diffusive, convective, and sedimentation mechanisms. Our study combines these techniques with CT scan simulation to study the mechanisms of respiratory movement and its imaging appearance. We use 3D fluid-structure interaction simulation to study the movement of an ideal alveolus under regular and forced breathing situations and ill alveoli with different tissue elasticities. Additionally, we use the Monte Carlo algorithm within the OpenGATE platform to simulate the computational CT images of the dynamic process with different designated resolutions. The resolutions show the relationship between the kinematic model of the human alveolus and its imaging appearance. The results show that the alveolus and the wall thickness can be seen with an image resolution smaller than 15.6μm. With ordinary CT resolution, the alveolus is expressed with four voxels. This is a preliminary study concerning the imaging appearance of the dynamic alveolus model. This technique will be used to study the imaging appearance of the dynamic bronchial tree and the lung lobe models in the future.

Developing an Online Interactive Volumetric Surface Viewer for Visualizing Spatial Distribution of Intracranial Lesions.

The exact location of intracranial neoplasms and metastatic lesions within the cortex may provide key insights into predicting treatment effectiveness and toxicities. Modern neuroimaging approaches have enabled precise, rapid diffeomorphic alignment from individual brains to an atlas, allowing the registration of contours and dose distributions onto a common template. However, there is a paucity of intuitive tools to perform such functions. Here, we developed an online interactive volumetric surface viewer that displays distribution of brain lesions of specific clinical features within the context of prior radiation treatment in populations of patients on a common atlas. Data from 647 patients treated at our institution with radiotherapy for intracranial lesions were analyzed with FreeSurfer to perform intensity normalization, skull-stripping, and affine registration to Talairach coordinates. A custom-built VoxelMorph model was trained to register T1 MRIs to a common atlas. Radiation dose maps and contours on underlying CT simulation scans were rigidly aligned to the T1 MRIs. The diffeomorphic warps were then applied to the radiation dose maps and contours, including the gross tumor volumes. Surface rendering of the template atlas was performed using FreeSurfer and visualized using PyCortex. Clinical information including primary sites for brain metastases, lesion size, survival, recurrence, use of anti-epileptic medications, radiation necrosis, and neuro-cognitive toxicity was embedded into the PyCortex viewer. The custom-trained VoxelMorph model enabled rapid normalization of the T1MRIs to the atlas brain. PyCortex allows for interactive visualization of the spatial distribution of lesions on the cortical surface across a rich array of clinical parameters. Initial maps have qualitatively revealed a predilection for metastases in the right hemisphere. This WebGL viewer allows intuitive clinical interrogation on a visual surface enabling population inferences. For example, one can visualize the distribution of brain metastases of a large patient cohort based on their primary tumor sites, propensity for seizure and risk for radionecrosis. We successfully built an online interactive viewer to visualize the spatial distribution of intracranial lesions on a common reference. Our approach extends prior work allowing visual interaction of lesions and real-time generation of inference maps by combining a front-end viewer and back-end database of clinical data. This viewer may represent a critical step towards uncovering the location-dependent breakdown of the blood-brain-barrier, differentiating the propensity of primary lesions metastasizing to brain regions, and allowing the discovery of associations of previously understudied spatial neighborhoods. We plan to release a publicly available web-based viewer to further explore, interrogate, and inform future radiation treatments.

Dosimetric Analysis of Proton Beam Therapy vs. Photon Radiotherapy for Cardiac Tumors with or without Deep Inspiratory Breath Holding: A Case Report

Proton beam therapy (PBT) has been demonstrated to deliver equivalent dosimetric radiation with the benefit of improved sparing of organs at risk (OAR). Deep inspiration breath holding (DIBH) is a commonly used method for reducing the radiation dose to the heart and lungs. However, few studies have ever reported the usage of DIBH combined with proton beam therapy in cardiac tumors. The purpose of this case report is to compare the dosimetric differences between photon radiotherapy and proton radiation therapy (PBT) with or without deep inspiration breath holding. A 66-year-old female patient with cardiac tumors was recruited, and the prescribed dose of radiotherapy for cardiac tumors was 95%PGTV 50Gy/2.5Gy/20f. Two simulation CT scans were collected during free breath (FB) and DIBH. And the target area was delineated on deep inspiratory breath holding image (DIBH-CT) and free breathing image (FB-CT). The target area of FB-CT was modified by referring to the ten-time phases of 4D-CT. Finally, IMRT, VMAT and PBT plans (DIBH-IMRT, DIBH-VAMT, DIBH-PBT, FB-IMRT, FB-VAMT, FB-PBT) were generated on the above images, and the organs at risk were limited as follows: lungs V20 ≤20%, lungs mean ≤11 Gy, heart V30 ≤40%, coronary artery mean ≤26 Gy, spinal cord ≤30 Gy, and left breast mean ≤5 Gy. All of the six plans satisfied most of the treatment planning goals. DIBH resulted in a dose reduction in all organs at risk including the heart, lungs, coronary artery (CA), spinal cord and breasts, when compared with FB using IMRT, VMAT, or PBT. Compared with the FB, DIBH provided a significant reduction in the mean dose of coronary artery (CA mean for DIBH-IMRT vs FB-IMRT = 28.32 Gy vs 42.66 Gy, CA mean for DIBH-VMAT vs FB-VAMT = 26.44Gy vs 40.85Gy, CA mean for DIBH-PBT vs FB-PBT = 27.71Gy vs 39.51Gy). Similarly, when compared with IMRT or VMAT in either FB or DIBH, PBT reduced radiation doses for all of the OAR. In comparison, the difference was less significant between IMRT and VMAT technique. Pitmen compared with IMRT and VMAT, reduced significantly the max dose of spinal cord, lungs V5, breast-L/R mean. Totally, DIBH-PBT was observed sufficient dose coverage and better sparing of organs at risk. PBT combined with DIBH technique gained an advantage in the sparing of OAR for cardiac tumors, especially in coronary protection. The possibility of broader application of PBT with DIBH in clinical practice is currently being evaluated and further studies are needed.

Non-Animal Stabilized Hyaluronic Acid (NASHA) Gel Marker vs. Surgical Clips for Tumor Bed Delineation in Breast Cancer Using MR-Simulation

Consistent delineation of the tumor bed (TB) after breast conserving surgery (BCS) is critical and remains a challenge with increasing prevalence of oncoplastic surgeries. Clips are generally used to help TB identification on CT-simulation but they are not well identified on MR-simulation. The aim of this study is to assess whether the use of NASHA gel with MR-simulation yields similar interobserver variability (inter-OV) compared to clips with CT-simulation. This prospective single arm feasibility study included patients eligible for BCS. After lumpectomy, the surgeon placed both clips (>5) and NASHA gel drops as markers to define the TB. Patients underwent CT and MRI simulation scans. Five radiation oncologists and one radiologist delineated the TB aided by clips on CT, and gel on MRI. The observers also assessed the visibility and utility of the gel (scale from 1 to 10), as well as the cavity visualization score (CVS, scale from 1 to 5). The primary endpoint was the inter-OV of the delineated TB using the overlap difference of contours using clips and CT versus gel and MRI, with the conformity index measured according to the pair definition of the Dice Similarity Coefficient (DSC). Of the 35 patients recruited, 30 were eligible for inter-OV analysis of TB delineation and 5 patients required further breast surgery for positive margins. One third of the eligible patients underwent an oncoplastic procedure. There were no significant differences between inter-OV of delineated TB using clips and CT versus gel and T2-weighted MRI with the mean DSC (0.60 vs 0.62, p = 0.364). The observers reported higher usefulness of gel in patients with an oncoplastic procedure than not (median US 8.2 vs 6.6, p = 0.024), and higher visibility of gel in patients who had their scans within 6 weeks than beyond post-op (median VS 8.1 vs 6.1, p = 0.013). When the CVS was higher (3-5), the median US of gel was lower (5.9 vs 7.8, p = 0.004), and the conformity index of clip and CT delineated TB was higher (median DSC 0.72 vs 0.53, p <0.001). Interestingly, a higher CVS did not lead to significantly higher conformity index of gel and T2-weighted MR delineated TB (mean DSC 0.67 vs 0.58, p = 0.073). NASHA gel injection added a median of 3 minutes to the operating theatre (OT) time and was rated as 'easy' in 89% of cases by surgeons. There were no immediate adverse events (AE) in OT, while 2 of 35 patients later experienced a grade 3 AE - hematoma which required evacuation in OT day 1 post-BCS, and infected seroma which required drainage and washout in OT 2 months post-BCS and axillary dissection. These reflect common risks with standard BCS and are not clearly attributed to gel injection alone. Use of NASHA gel leads to similar inter-OV of BC TB delineation compared to >5 clips. NASHA gel is hence a reliable alternative to clips when MR-simulation is used.

The Feasibility of Adaptive CT Simulation-Free Radiation Therapy

To investigate the feasibility of a novel, streamlined radiotherapy treatment workflow in which simulation and treatment occur in a single on-table session. The efficiency, dosimetric accuracy, plan quality, and resource requirements were metrics used to evaluate the clinical potential of CT simulation-free radiotherapy. An adaptive radiotherapy treatment platform was upgraded to include a novel CBCT system reporting image quality and Hounsfield unit accuracy specifications comparable to standard fan-beam CT. An in-house workflow was designed to allow the CBCT data to be used as reference image sets for planning. Two test cases were completed on anthropomorphic phantoms equipped with small volume ion chambers (cross-calibrated to an ADCL traceable dose standard) to evaluate the feasibility and accuracy of the workflows. A spine was planned for palliative treatment with 8 Gy in 1 fraction, and an intact prostate was planned for definitive treatment with 60 Gy in 20 fractions. The CBCTs were acquired using default thorax and pelvis imaging protocols and reconstructed using an iterative algorithm with scatter removal. CBCTs were used for contouring and planning, and treatment was delivered via an online adaptive workflow. An extended-field CBCT acquisition can be acquired in 12 s of acquisition time and reconstructed in approximately 1 min. The superior-inferior extent for the CBCT planning images was 42 cm, which captured the full extent of relevant anatomy; an audit of recent single-site palliative CT simulation scans had an average length of 44.9 cm. The contouring and treatment planning for the spine and prostate were completed in 30 min and 23 min, respectively. The dosimetric agreement with the treatment plan agreed to within a range of -1.17 to 1.64%, and a mean and standard deviation of 0.41 ± 1.16%. Using an in-house workflow, CT simulation-free radiation therapy was shown to be feasible with acceptable workflow efficiency and dosimetric accuracy. This approach may be particularly applicable for urgent palliative treatments. With the availability of software to enable this workflow, and the continued advancement of on-treatment adaptation, single-visit radiation therapy may replace current practice for some clinical indications.

The Role of Coronary Artery Calcium Score to Assess Risk of Cardiovascular Disease in Irradiated Esophageal Cancer Patients

Coronary artery calcium (CAC) score is an important predictive imaging marker of cardiovascular disease (CVD). While studies have found positive association between CAC score and cardiac toxicity in irradiated lung and breast cancer patients, there are no studies assessing CAC scores in esophageal cancer (EC). While a cardiac-gated CT is required for standard Agatston CAC score, visual assessment of CAC via ordinal scoring on non-gated CT has shown good concordance with Agatston score. In this study, we sought to examine whether visual assessment of CAC, measured on standard of care, non-contrast chest CT, predicts the development of adverse cardiovascular events (ACVE) in irradiated EC patients. This is a single institution retrospective study of EC patients treated with RT from 2010-2021. We included patients with available PET/CT at diagnosis or chest CT simulation scan without contrast, and excluded those with history of percutaneous coronary intervention, coronary bypass surgery, or prior thoracic RT. Pre-treatment characteristics, clinical factors, and grade ≥ 3 (G3+) adverse cardiovascular events (ACVE) (CTCAEv5.0) were evaluated. Visual assessment of CAC was performed using ordinal method (CAC scored from 0 to 12), by a thoracic radiologist. Fine and Gray regression was used to compute hazard ratios for time to first ACVE. Univariate analyses using Cox proportional hazards were used for overall survival (OS). ACVEs were recorded from start of oncologic treatment and OS calculated after completion of RT. A total of 118 patients were analyzed with a median follow-up of 16 months. Median age was 67 years, 65% male, 43% white, 59% with EC of distal esophagus, and 59% had squamous cell carcinoma. Median mean heart dose was 21.93 Gy (range 0.15-36.94). 24% developed G3+ ACVEs: atrial fibrillation 9%, stroke 6%, heart failure 4%, pulmonary embolism 4%, pericardial effusion 3%, myocardial infarction 2%, heart block 2%, and cardiac death 1%. On univariate analyses, CAC >1 vs. CAC ≤ 1 trended towards increased risk of ACVE (HR = 1.95, 95% CI = 0.89-4.26; p = 0.094), however it is not predictive of OS (HR = 1.31, 95% CI = 0.75-2.30; p = 0.343). Proportion of patients with ACVEs was greater in CAC>1 group (Table). When compared to patients with CAC ≤ 1, those with CAC >1 were older (median age 62 vs 72 years, p = 0.0015), less likely to be never smokers (38% vs 30%, p = 0.0437), and more likely to have hypertension (43% vs 64%, p = 0.0197), and hyperlipidemia (30% vs 47%, p = 0.0557). This is the first study to investigate the relationship between CAC score and ACVEs in EC. While the study was underpowered (likely due to low rates of recorded ACVEs), to detect a significant association between CAC score and ACVEs, there was a trend towards increased risk of ACVEs in patients with a CAC score >1 by visual ordinal scoring. Further prospective evaluation with a larger cohort is warranted.

Failure Mode and Effects Analysis Prior to the Introduction of AI Generated GTVs for Brain Metastases in the Clinical Workflow

AI autosegmentation of organs-at-risk (OARs) is common practice at many radiotherapy clinics. Despite the abundance of gross tumor volume (GTV) autosegmentation algorithms, adoption in clinical care has been slow due to the high risk associated with errors in GTV delineation. Here we present a failure mode and effects analysis (FMEA) to evaluate the risk associated with introducing AI derived GTVs in patients treated with stereotactic radiosurgery (SRS). An AI GTV autosegmentation algorithm for brain metastases was developed in-house based on a V-Net 3D CNN. Registered CT and MR images and a contour of the brain are input into the software and all identified lesions are returned in a DICOM-RT structure set. Following algorithm evaluation, a workflow was developed to enable AI GTV autosegmentation to be introduced clinically for every SRS patient. The following steps were added to existing procedures: 1) workflow to send CT/MR and brain structure to external server, 2) autosegmentation run on the server, 3) AI GTV structures with a standard nomenclature added to existing OAR structure set, and 4) MD review, editing, and approval of AI GTVs. After successfully completing the physics evaluation testing of the new process, we formed a team of 10 faculty and staff including physicists, residents, physicians, and planners to perform the FMEA prior to clinical implementation. The team met to map the process, identify potential failure modes, and score their frequency of occurrence, severity, and detectability. A 3-point scale (1, 3, or 5) was used to simplify the scoring process. Occurrence was defined as rare, sometimes, or often; severity as low, medium, or high; and detectability as obvious, possible, or challenging. The risk probability numbers (RPNs) were calculated and the steps in the process with the highest RPNs were flagged for further discussion. The FMEA team completed their process map and analysis primarily in 4 meetings. The process map began with acquisition of the patients CT simulation scan and ended with physician approval of final volumes for treatment planning. We identified 17 process steps and 72 possible failure modes, of which 26 were associated with the new workflow. Eighteen failure modes had an RPN greater than 30 (highest risk score in at least one category) and were flagged to assess mitigation strategies. Five were unique to the new AI GTV workflow and mitigation strategies will be designed prior to clinical use. Those involved risks related to inaccurate AI GTV contours, false positives, and an incomplete review stemming from over-reliance by team members on AI. AI is increasingly being employed at every step of radiotherapy to automate and streamline processes. The FMEA analysis resulted in the identification of the riskiest parts of using AI GTV autosegmentation. This can be an effective tool in the development of checks to ensure that GTV autosegmentation methods can be safely introduced in support of patient care.

Predicting successful clinical candidates for fiducial-free lung tumor tracking with a deep learning binary classification model.

The CyberKnife system is a robotic radiosurgery platform that allows the delivery of lung SBRT treatments using fiducial-free soft-tissue tracking. However, not all lung cancer patients are eligible for lung tumor tracking. Tumor size, density, and location impact the ability to successfully detect and track a lung lesion in 2D orthogonal X-ray images. The standard workflow to identify successful candidates for lung tumor tracking is called Lung Optimized Treatment (LOT) simulation, and involves multiple steps from CT acquisition to the execution of the simulation plan on CyberKnife. The aim of the study is to develop a deep learning classification model to predict which patients can be successfully treated with lung tumor tracking, thus circumventing the LOT simulationprocess. Target tracking is achieved by matching orthogonal X-ray images with a library of digital radiographs reconstructed from the simulation CT scan (DRRs). We developed a deep learning model to create a binary classification of lung lesions as being trackable or untrackable based on tumor template DRR extracted from the CyberKnife system, and tested five different network architectures. The study included a total of 271 images (230 trackable, 41 untrackable) from 129 patients with one or multiple lung lesions. Eighty percent of the images were used for training, 10% for validation, and the remaining 10% for testing. For all five convolutional neural networks, the binary classification accuracy reached 100% after training, both in the validation and the test set, without any falseclassifications. A deep learning model can distinguish features of trackable and untrackable lesions in DRR images, and can predict successful candidates for fiducial-free lung tumortracking.

P17.17.A WHAT IS THE BETTER THERAPEUTIC STRATEGY FOR DISTANT RECURRENT GLIOMA? OUTCOME AND PROGNOSTIC FACTORS EVALUATION IN A RETROSPECTIVE SERIES

Abstract BACKGROUND Distant recurrent glioma is a rare condition. A possible correlation has been postulate with original tumor location in close proximity to the subventricular zone. Nevertheless, to date no data exist about treatment strategy, neither retrospective nor prospective studies, and a standard of care is missing. Several treatment options are available: surgery, systemic treatment (CHT), and radiation therapy (RT) alone or with concomitant chemotherapy. Aim of this retrospective analysis was to investigate outcome of patients (pts) with distant recurrent glioma treated with RT, with or without concomitant and/or adjuvant temozolomide chemotherapy. MATERIAL AND METHODS Adult pts with distant recurrent glioma have been evaluated. Histological diagnosis has been reclassified according to the 2021 WHO Classification of Central Nervous System tumors, and original tumor proximity to subventricular zone has been registered. Distant progression treatment strategies included RT alone, RT concurrent with CHT (RT-CHT), RT followed by CHT. For RT planning, target delineation was defined on simulation CT scan and brain MRI co-registered into the image workstation of Brainlab. RESULTS From 2015 to 2023, 35 distant recurrent glioma pts have been evaluated. Histological diagnosis was: glioblastoma (GBM) in 27 pts, astrocytoma grade 3 IDH-mutant in 6 pts, and astrocytoma grade 2 IDH-mutant in 2 pts. Location of primary tumor was in subventricular zone in all cases: 16 pts developed only distant recurrence, 19 pts both local and distant recurrence. Median interval time (IT) between diagnosis and distant recurrence was 22 months (range 5-49); 8 cases had distant recurrence within 12 months, 17 cases between 13 and 24 months, and 10 cases ≥ 24 months. Regarding treatment strategy, 10 pts (29%) received RT alone, 6 pts (17%) received RT-CHT, 11 (31%) received RT followed by sequential CHT, 8 (23%) received concurrent RT-CHT followed by sequential CHT. Median RT dose was 45 Gy in 15 fractions (range 30-50 Gy in 5-20fr). Median follow up time was 36months (range3-97). Median overall survival (OS) time, 6 months,1,2-year OS rates from diagnosis were 33 months (95% CI 30-36), 100%, 94.3%±3.92, 70.0%±7.9, respectively. Median OS time, 6 months,1,2-year OS rates from distant recurrence were 10 months (95% CI 7-12), 75.4%±7.6, 26.2%±7.9, 0.%. On univariate, and multivariate analysis, prognostic factors impacting OS were IT between diagnosis and first recurrence, and treatment performed. Indeed, median OS time were 6 months (95% CI 2-10) for IT ≤ 1year (P= 0.0363), 17 months (95% CI 8-12), 10 months (95% CI 7-12), 6 months (95% CI 8-12) for RT alone, RT and concurrent RT-CHT followed by sequential CHT, concurrent RT-CHT without sequential CHT, respectively (P &amp;lt; 0.0001). CONCLUSION Our results suggest that RT alone could be a feasible and effective treatment option for distant recurrent glioma.

Adipose Tissue Distribution and Body Mass Index (BMI) Correlation With Daily Image-Guided Radiotherapy (IGRT) Shifts of Abdominal Radiation Therapy Patients.

Purpose There are several studies suggesting a correlation between image-guided radiotherapy (IGRT) setup errors and body mass index (BMI). However, abdominal fat content has visceral and subcutaneous components, which may affect setup errors differently. This study aims to analyze a potential workflow for characterizing adipose content and distribution in the region of the target that would allow a quickly calculated metric of abdominal fat content to stratify these patients. Methods IGRT shift data was retrospectively tabulated from daily fan-beam CT-on-rails pre-treatment alignment for 50 abdominal radiation therapy (RT) patients, and systematic and random errors in the daily setup were characterized by tabulating average and standard deviations of shift data for each patient and looking at differences for different distributions of adipose content. Visceral and subcutaneous fat content were defined by visceral fat area (VFA) and subcutaneous fat area (SFA) using a region-growing algorithm to contour adipose tissue on CT simulation scans. All contours were created for a single slice at the treatment isocenter, on which the VFA and SFA were calculated. A log-rank test was used to test trends in shifts over quartiles of adiposity. Results VFA ranged from 1.9-342.8c m2, and SFA from 11.8-756.0 cm2. The standard definition (SD) of random error (σ) in the lateral axis for Q1 vs. Q4 VFA was 0.10cm vs. 0.29cm, 0.12cm vs. 0.28cm for SFA, and 0.12cm vs. 0.31cm for BMI. The percentage of longitudinal shifts greater than 10mm for Q1 vs. Q4 VFA was 0% vs. 9%, 2% vs. 19% for SFA, and 0% vs. 20% for BMI. Statistically significant trends in shifts vs. the BMI quartile were seen for both pitch and the longitudinal direction, as well as for pitch corrections vs. the VFA quartile. Conclusion Within this dataset, abdominal cancer patients showed statistically significant trends in shift probability vs. BMI and VFA. Also, patients in the upper quartiles of all adiposity metrics showed an increased SD of σ in the lateral direction and increased shifts over 10 mm in the longitudinal direction. However, despite these relationships, neither VFA nor SFA offered discernible advantages in their relationship to shift uncertainty relative to BMI.

Bony structure enhanced synthetic CT generation using Dixon sequences for pelvis MR-only radiotherapy.

MRI-only radiotherapy planning (MROP) is beneficial to patients by avoiding MRI/CT registration errors, simplifying the radiation treatment simulation workflow and reducing exposure to ionizing radiation. MRI is the primary imaging modality for soft tissue delineation. Treatment planning CTs (i.e., CT simulation scan) are redundant if a synthetic CT (sCT) can be generated from the MRI to provide the patient positioning and electron density information. Unsupervised deep learning (DL) models like CycleGAN are widely used in MR-to-sCT conversion, when paired patient CT and MR image datasets are not available for model training. However, compared to supervised DL models, they cannot guarantee anatomic consistency, especially around bone. The purpose of this work was to improve the sCT accuracy generated from MRI around bone for MROP. To generate more reliable bony structures on sCT images, we proposed to add bony structure constraints in the unsupervised CycleGAN model's loss function and leverage Dixon constructed fat and in-phase (IP) MR images. Dixon images provide better bone contrast than T2-weighted images as inputs to a modified multi-channel CycleGAN. A private dataset with a total of 31 prostate cancer patients were used for training (20) and testing (11). We compared model performance with and without bony structure constraints using single- and multi-channel inputs. Among all the models, multi-channel CycleGAN with bony structure constraints had the lowest mean absolute error, both inside the bone and whole body (50.7 and 145.2 HU). This approach also resulted in the highest Dice similarity coefficient (0.88) of all bony structures compared with the planning CT. Modified multi-channel CycleGAN with bony structure constraints, taking Dixon-constructed fat and IP images as inputs, can generate clinically suitable sCT images in both bone and soft tissue. The generated sCT images have the potential to be used for accurate dose calculation and patient positioning in MROP radiation therapy.