Evaluating automatically generated normal tissue contours for safe use in head and neck and cervical cancer treatment planning.

PURPOSE: Volumetric-modulated arc therapy (VMAT) is a widely accepted treatment method for head and neck (HN) and cervical cancers; however, creating contours and plan optimization for VMAT plans is a time-consuming process. Our group has created an automated treatment planning tool, the Radiation Planning Assistant (RPA), that uses deep learning models to generate organs at risk (OARs), planning structures and automates plan optimization. This study quantitatively evaluates the quality of contours generated by the RPA tool. METHODS: For patients with HN (54) and cervical (39) cancers, we retrospectively generated autoplans using the RPA. Autoplans were generated using deep-learning and RapidPlan models developed in-house. The autoplans were, then, applied to the original, physician-drawn contours, which were used as a ground truth (GT) to compare with the autocontours (RPA). Using a "two one-sided tests" (TOST) procedure, we evaluated whether the autocontour normal tissue dose was equivalent to that of the ground truth by a margin, δ, that we determined based on clinical judgement. We also calculated the number of plans that met established clinically accepted dosimetric criteria. RESULTS: For HN plans, 91.8% and 91.7% of structures met dosimetric criteria for automatic and manual contours, respectively; for cervical plans, 95.6% and 95.7% of structures met dosimetric criteria for automatic and manual contours, respectively. Autocontours were equivalent to the ground truth for 71% and 75% of common DVH metrics for the HN and cervix, respectively. CONCLUSIONS: This study shows that dosimetrically equivalent normal tissue contours can be created for HN and cervical cancers using deep learning techniques. In general, differences between the contours did not affect the passing or failing of clinical dose tolerances.

Random Forest Modeling of Acute Toxicity in Anal Cancer: Effects of Peritoneal Cavity Contouring Approaches on Model Performance.

PURPOSE: Our purpose was to analyze the effect on gastrointestinal (GI) toxicity models when their dose-volume metrics predictors are derived from segmentations of the peritoneal cavity after different contouring approaches. METHODS AND MATERIALS: A random forest machine learning approach was used to predict acute grade ≥3 GI toxicity from dose-volume metrics and clinicopathologic factors for 246 patients (toxicity incidence = 9.5%) treated with definitive chemoradiation for squamous cell carcinoma of the anus. Three types of random forest models were constructed based on different bowel bag segmentation approaches: (1) physician-delineated after Radiation Therapy Oncology Group (RTOG) guidelines, (2) autosegmented by a deep learning model (nnU-Net) following RTOG guidelines, and (3) autosegmented but spanning the entire bowel space. Each model type was evaluated using repeated cross-validation (100 iterations; 50%/50% training/test split). The performance of the models was assessed using area under the precision-recall curve (AUPRC) and the receiver operating characteristic curve (AUROCC), as well as optimal F1 score. RESULTS: When following RTOG guidelines, the models based on the nnU-Net auto segmentations (mean values: AUROCC, 0.71 ± 0.07; AUPRC, 0.42 ± 0.09; F1 score, 0.46 ± 0.08) significantly outperformed (P < .001) those based on the physician-delineated contours (mean values: AUROCC, 0.67 ± 0.07; AUPRC, 0.34 ± 0.08; F1 score, 0.36 ± 0.07). When spanning the entire bowel space, the performance of the autosegmentation models improved considerably (mean values: AUROCC, 0.87 ± 0.05; AUPRC, 0.70 ± 0.09; F1 score, 0.68 ± 0.09). CONCLUSIONS: Random forest models were superior at predicting acute grade ≥3 GI toxicity when based on RTOG-defined bowel bag autosegmentations rather than physician-delineated contours. Models based on autosegmentations spanning the entire bowel space show further considerable improvement in model performance. The results of this study should be further validated using an external data set.

Deep learning-based automatic segmentation of cardiac substructures for lung cancers.

PURPOSE: Accurate and comprehensive segmentation of cardiac substructures is crucial for minimizing the risk of radiation-induced heart disease in lung cancer radiotherapy. We sought to develop and validate deep learning-based auto-segmentation models for cardiac substructures. MATERIALS AND METHODS: Nineteen cardiac substructures (whole heart, 4 heart chambers, 6 great vessels, 4 valves, and 4 coronary arteries) in 100 patients treated for non-small cell lung cancer were manually delineated by two radiation oncologists. The valves and coronary arteries were delineated as planning risk volumes. An nnU-Net auto-segmentation model was trained, validated, and tested on this dataset with a split ratio of 75:5:20. The auto-segmented contours were evaluated by comparing them with manually drawn contours in terms of Dice similarity coefficient (DSC) and dose metrics extracted from clinical plans. An independent dataset of 42 patients was used for subjective evaluation of the auto-segmentation model by 4 physicians. RESULTS: The average DSCs were 0.95 (+/- 0.01) for the whole heart, 0.91 (+/- 0.02) for 4 chambers, 0.86 (+/- 0.09) for 6 great vessels, 0.81 (+/- 0.09) for 4 valves, and 0.60 (+/- 0.14) for 4 coronary arteries. The average absolute errors in mean/max doses to all substructures were 1.04 (+/- 1.99) Gy and 2.20 (+/- 4.37) Gy. The subjective evaluation revealed that 94% of the auto-segmented contours were clinically acceptable. CONCLUSION: We demonstrated the effectiveness of our nnU-Net model for delineating cardiac substructures, including coronary arteries. Our results indicate that this model has promise for studies regarding radiation dose to cardiac substructures.

Synthetic Megavoltage Cone Beam Computed Tomography Image Generation for Improved Contouring Accuracy of Cardiac Pacemakers.

In this study, we aimed to enhance the contouring accuracy of cardiac pacemakers by improving their visualization using deep learning models to predict MV CBCT images based on kV CT or CBCT images. Ten pacemakers and four thorax phantoms were included, creating a total of 35 combinations. Each combination was imaged on a Varian Halcyon (kV/MV CBCT images) and Siemens SOMATOM CT scanner (kV CT images). Two generative adversarial network (GAN)-based models, cycleGAN and conditional GAN (cGAN), were trained to generate synthetic MV (sMV) CBCT images from kV CT/CBCT images using twenty-eight datasets (80%). The pacemakers in the sMV CBCT images and original MV CBCT images were manually delineated and reviewed by three users. The Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), and mean surface distance (MSD) were used to compare contour accuracy. Visual inspection showed the improved visualization of pacemakers on sMV CBCT images compared to original kV CT/CBCT images. Moreover, cGAN demonstrated superior performance in enhancing pacemaker visualization compared to cycleGAN. The mean DSC, HD95, and MSD for contours on sMV CBCT images generated from kV CT/CBCT images were 0.91 ± 0.02/0.92 ± 0.01, 1.38 ± 0.31 mm/1.18 ± 0.20 mm, and 0.42 ± 0.07 mm/0.36 ± 0.06 mm using the cGAN model. Deep learning-based methods, specifically cycleGAN and cGAN, can effectively enhance the visualization of pacemakers in thorax kV CT/CBCT images, therefore improving the contouring precision of these devices.

Radiation Planning Assistant - A Web-based Tool to Support High-quality Radiotherapy in Clinics with Limited Resources.

Access to radiotherapy worldwide is limited. The Radiation Planning Assistant (RPA) is a fully automated, web-based tool that is being developed to offer fully automated radiotherapy treatment planning tools to clinics with limited resources. The goal is to help clinical teams scale their efforts, thus reaching more patients with cancer. The user connects to the RPA via a webpage, completes a Service Request (prescription and information about the radiotherapy targets), and uploads the patient's CT image set. The RPA offers two approaches to automated planning. In one-step planning, the system uses the Service Request and CT scan to automatically generate the necessary contours and treatment plan. In two-step planning, the user reviews and edits the automatically generated contours before the RPA continues to generate a volume-modulated arc therapy plan. The final plan is downloaded from the RPA website and imported into the user's local treatment planning system, where the dose is recalculated for the locally commissioned linac; if necessary, the plan is edited prior to approval for clinical use.

Automated contouring and statistical process control for plan quality in a breast clinical trial.

Background and purpose: Automatic review of breast plan quality for clinical trials is time-consuming and has some unique challenges due to the lack of target contours for some planning techniques. We propose using an auto-contouring model and statistical process control to independently assess planning consistency in retrospective data from a breast radiotherapy clinical trial. Materials and methods: A deep learning auto-contouring model was created and tested quantitatively and qualitatively on 104 post-lumpectomy patients' computed tomography images (nnUNet; train/test: 80/20). The auto-contouring model was then applied to 127 patients enrolled in a clinical trial. Statistical process control was used to assess the consistency of the mean dose to auto-contours between plans and treatment modalities by setting control limits within three standard deviations of the data's mean. Two physicians reviewed plans outside the limits for possible planning inconsistencies. Results: Mean Dice similarity coefficients comparing manual and auto-contours was above 0.7 for breast clinical target volume, supraclavicular and internal mammary nodes. Two radiation oncologists scored 95% of contours as clinically acceptable. The mean dose in the clinical trial plans was more variable for lymph node auto-contours than for breast, with a narrower distribution for volumetric modulated arc therapy than for 3D conformal treatment, requiring distinct control limits. Five plans (5%) were flagged and reviewed by physicians: one required editing, two had clinically acceptable variations in planning, and two had poor auto-contouring. Conclusions: An automated contouring model in a statistical process control framework was appropriate for assessing planning consistency in a breast radiotherapy clinical trial.

A real-time contouring feedback tool for consensus-based contour training.

Purpose: Variability in contouring structures of interest for radiotherapy continues to be challenging. Although training can reduce such variability, having radiation oncologists provide feedback can be impractical. We developed a contour training tool to provide real-time feedback to trainees, thereby reducing variability in contouring. Methods: We developed a novel metric termed localized signed square distance (LSSD) to provide feedback to the trainee on how their contour compares with a reference contour, which is generated real-time by combining trainee contour and multiple expert radiation oncologist contours. Nine trainees performed contour training by using six randomly assigned training cases that included one test case of the heart and left ventricle (LV). The test case was repeated 30 days later to assess retention. The distribution of LSSD maps of the initial contour for the training cases was combined and compared with the distribution of LSSD maps of the final contours for all training cases. The difference in standard deviations from the initial to final LSSD maps, ΔLSSD, was computed both on a per-case basis and for the entire group. Results: For every training case, statistically significant ΔLSSD were observed for both the heart and LV. When all initial and final LSSD maps were aggregated for the training cases, before training, the mean LSSD ([range], standard deviation) was -0.8 mm ([-37.9, 34.9], 4.2) and 0.3 mm ([-25.1, 32.7], 4.8) for heart and LV, respectively. These were reduced to -0.1 mm ([-16.2, 7.3], 0.8) and 0.1 mm ([-6.6, 8.3], 0.7) for the final LSSD maps during the contour training sessions. For the retention case, the initial and final LSSD maps of the retention case were aggregated and were -1.5 mm ([-22.9, 19.9], 3.4) and -0.2 mm ([-4.5, 1.5], 0.7) for the heart and 1.8 mm ([-16.7, 34.5], 5.1) and 0.2 mm ([-3.9, 1.6],0.7) for the LV. Conclusions: A tool that uses real-time contouring feedback was developed and successfully used for contour training of nine trainees. In all cases, the utility was able to guide the trainee and ultimately reduce the variability of the trainee's contouring.

Addressing the Global Expertise Gap in Radiation Oncology: The Radiation Planning Assistant.

PURPOSE: Automation, including the use of artificial intelligence, has been identified as a possible opportunity to help reduce the gap in access and quality for radiotherapy and other aspects of cancer care. The Radiation Planning Assistant (RPA) project was conceived in 2015 (and funded in 2016) to use automated contouring and treatment planning algorithms to support the efforts of oncologists in low- and middle-income countries, allowing them to scale their efforts and treat more patients safely and efficiently (to increase access). DESIGN: In this review, we discuss the development of the RPA, with a particular focus on clinical acceptability and safety/risk across jurisdictions as these are important indicators for the successful future deployment of the RPA to increase radiotherapy availability and ameliorate global disparities in access to radiation oncology. RESULTS: RPA tools will be offered through a webpage, where users can upload computed tomography data sets and download automatically generated contours and treatment plans. All interfaces have been designed to maximize ease of use and minimize risk. The current version of the RPA includes automated contouring and planning for head and neck cancer, cervical cancer, breast cancer, and metastases to the brain. CONCLUSION: The RPA has been designed to bring high-quality treatment planning to more patients across the world, and it may encourage greater investment in treatment devices and other aspects of cancer treatment.

Deep Learning-Based Dose Prediction for Automated, Individualized Quality Assurance of Head and Neck Radiation Therapy Plans.

PURPOSE: This study aimed to use deep learning-based dose prediction to assess head and neck (HN) plan quality and identify suboptimal plans. METHODS AND MATERIALS: A total of 245 volumetric modulated arc therapy HN plans were created using RapidPlan knowledge-based planning (KBP). A subset of 112 high-quality plans was selected under the supervision of an HN radiation oncologist. We trained a 3D Dense Dilated U-Net architecture to predict 3-dimensional dose distributions using 3-fold cross-validation on 90 plans. Model inputs included computed tomography images, target prescriptions, and contours for targets and organs at risk (OARs). The model's performance was assessed on the remaining 22 test plans. We then tested the application of the dose prediction model for automated review of plan quality. Dose distributions were predicted on 14 clinical plans. The predicted versus clinical OAR dose metrics were compared to flag OARs with suboptimal normal tissue sparing using a 2 Gy dose difference or 3% dose-volume threshold. OAR flags were compared with manual flags by 3 HN radiation oncologists. RESULTS: The predicted dose distributions were of comparable quality to the KBP plans. The differences between the predicted and KBP-planned D1%,D95%, and D99% across the targets were within -2.53% ± 1.34%, -0.42% ± 1.27%, and -0.12% ± 1.97%, respectively, and the OAR mean and maximum doses were within -0.33 ± 1.40 Gy and -0.96 ± 2.08 Gy, respectively. For the plan quality assessment study, radiation oncologists flagged 47 OARs for possible plan improvement. There was high interphysician variability; 83% of physician-flagged OARs were flagged by only one of 3 physicians. The comparative dose prediction model flagged 63 OARs, including 30 of 47 physician-flagged OARs. CONCLUSIONS: Deep learning can predict high-quality dose distributions, which can be used as comparative dose distributions for automated, individualized assessment of HN plan quality.

Automating the treatment planning process for 3D-conformal pediatric craniospinal irradiation therapy.

PURPOSE: Pediatric patients with medulloblastoma in low- and middle-income countries (LMICs) are most treated with 3D-conformal photon craniospinal irradiation (CSI), a time-consuming, complex treatment to plan, especially in resource-constrained settings. Therefore, we developed and tested a 3D-conformal CSI autoplanning tool for varying patient lengths. METHODS AND MATERIALS: Autocontours were generated with a deep learning model trained:tested (80:20 ratio) on 143 pediatric medulloblastoma CT scans (patient ages: 2-19 years, median = 7 years). Using the verified autocontours, the autoplanning tool generated two lateral brain fields matched to a single spine field, an extended single spine field, or two matched spine fields. Additional spine subfields were added to optimize the corresponding dose distribution. Feathering was implemented (yielding nine to 12 fields) to give a composite plan. Each planning approach was tested on six patients (ages 3-10 years). A pediatric radiation oncologist assessed clinical acceptability of each autoplan. RESULTS: The autocontoured structures' average Dice similarity coefficient ranged from .65 to .98. The average V95 for the brain/spinal canal for single, extended, and multi-field spine configurations was 99.9% ± 0.06%/99.9% ± 0.10%, 99.9% ± 0.07%/99.4% ± 0.30%, and 99.9% ± 0.06%/99.4% ± 0.40%, respectively. The average maximum dose across all field configurations to the brainstem, eyes (L/R), lenses (L/R), and spinal cord were 23.7 ± 0.08, 24.1 ± 0.28, 13.3 ± 5.27, and 25.5 ± 0.34 Gy, respectively (prescription = 23.4 Gy/13 fractions). Of the 18 plans tested, all were scored as clinically acceptable as-is or clinically acceptable with minor, time-efficient edits preferred or required. No plans were scored as clinically unacceptable. CONCLUSION: The autoplanning tool successfully generated pediatric CSI plans for varying patient lengths in 3.50 ± 0.4 minutes on average, indicating potential for an efficient planning aid in a resource-constrained settings.

Customizable landmark-based field aperture design for automated whole-brain radiotherapy treatment planning.

PURPOSE: To develop and evaluate an automated whole-brain radiotherapy (WBRT) treatment planning pipeline with a deep learning-based auto-contouring and customizable landmark-based field aperture design. METHODS: The pipeline consisted of the following steps: (1) Auto-contour normal structures on computed tomography scans and digitally reconstructed radiographs using deep learning techniques, (2) locate the landmark structures using the beam's-eye-view, (3) generate field apertures based on eight different landmark rules addressing different clinical purposes and physician preferences. Two parallel approaches for generating field apertures were developed for quality control. The performance of the generated field shapes and dose distributions were compared with the original clinical plans. The clinical acceptability of the plans was assessed by five radiation oncologists from four hospitals. RESULTS: The performance of the generated field apertures was evaluated by the Hausdorff distance (HD) and mean surface distance (MSD) from 182 patients' field apertures used in the clinic. The average HD and MSD for the generated field apertures were 16 ± 7 and 7 ± 3 mm for the first approach, respectively, and 17 ± 7 and 7 ± 3 mm, respectively, for the second approach. The differences regarding HD and MSD between the first and the second approaches were 1 ± 2 and 1 ± 3 mm, respectively. A clinical review of the field aperture design, conducted using 30 patients, achieved a 100% acceptance rate for both the first and second approaches, and the plan review achieved a 100% acceptance rate for the first approach and a 93% acceptance rate for the second approach. The average acceptance rate for meeting lens dosimetric recommendations was 80% (left lens) and 77% (right lens) for the first approach, and 70% (both left and right lenses) for the second approach, compared with 50% (left lens) and 53% (right lens) for the clinical plans. CONCLUSION: This study provided an automated pipeline with two field aperture generation approaches to automatically generate WBRT treatment plans. Both quantitative and qualitative evaluations demonstrated that our novel pipeline was comparable with the original clinical plans.

Automated Brain Metastases Segmentation With a Deep Dive Into False-positive Detection.

Purpose: The clinical management of brain metastases after stereotactic radiosurgery (SRS) is difficult, because a physician must review follow-up magnetic resonance imaging (MRI) scans to determine treatment outcome, which is often labor intensive. The purpose of this study was to develop an automated framework to contour brain metastases in MRI to help treatment planning for SRS and understand its limitations. Methods and Materials: Two self-adaptive nnU-Net models trained on postcontrast 3-dimensional T1-weighted MRI scans from patients who underwent SRS were analyzed. Performance was evaluated by computing positive predictive value (PPV), sensitivity, and Dice similarity coefficient (DSC). The training and testing sets included 3482 metastases on 845 patient MRI scans and 930 metastases on 206 patient MRI scans, respectively. Results: In the per-patient analysis, PPV was 90.1% ± 17.7%, sensitivity 88.4% ± 18.0%, DSC 82.2% ± 9.5%, and false positive (FP) 0.4 ± 1.0. For large metastases (≥6 mm), the per-patient PPV was 95.6% ± 17.5%, sensitivity 94.5% ± 18.1%, DSC 86.8% ± 7.5%, and FP 0.1 ± 0.4. The quality of autosegmented true-positive (TP) contours was also assessed by 2 physicians using a 5-point scale for clinical acceptability. Seventy-five percent of contours were assigned scores of 4 or 5, which shows that contours could be used as-is in clinical application, and the remaining 25% were assigned a score of 3, which means they needed minor editing only. Notably, a deep dive into FPs indicated that 9% were TP metastases not identified on the original radiology review, but identified on subsequent follow-up imaging (early detection). Fifty-four percent were real metastases (TP) that were identified but purposefully not contoured for target treatment, mainly because the patient underwent whole-brain radiation therapy before/after SRS treatment. Conclusions: These findings show that our tool can help radiologists and radiation oncologists detect and contour tumors from MRI, make precise decisions about suspicious lesions, and potentially find lesions at early stages.

Multi-organ segmentation of abdominal structures from non-contrast and contrast enhanced CT images.

Manually delineating upper abdominal organs at risk (OARs) is a time-consuming task. To develop a deep-learning-based tool for accurate and robust auto-segmentation of these OARs, forty pancreatic cancer patients with contrast-enhanced breath-hold computed tomographic (CT) images were selected. We trained a three-dimensional (3D) U-Net ensemble that automatically segments all organ contours concurrently with the self-configuring nnU-Net framework. Our tool's performance was assessed on a held-out test set of 30 patients quantitatively. Five radiation oncologists from three different institutions assessed the performance of the tool using a 5-point Likert scale on an additional 75 randomly selected test patients. The mean (± std. dev.) Dice similarity coefficient values between the automatic segmentation and the ground truth on contrast-enhanced CT images were 0.80 ± 0.08, 0.89 ± 0.05, 0.90 ± 0.06, 0.92 ± 0.03, 0.96 ± 0.01, 0.97 ± 0.01, 0.96 ± 0.01, and 0.96 ± 0.01 for the duodenum, small bowel, large bowel, stomach, liver, spleen, right kidney, and left kidney, respectively. 89.3% (contrast-enhanced) and 85.3% (non-contrast-enhanced) of duodenum contours were scored as a 3 or above, which required only minor edits. More than 90% of the other organs' contours were scored as a 3 or above. Our tool achieved a high level of clinical acceptability with a small training dataset and provides accurate contours for treatment planning.

VerSe: A Vertebrae labelling and segmentation benchmark for multi-detector CT images.

Vertebral labelling and segmentation are two fundamental tasks in an automated spine processing pipeline. Reliable and accurate processing of spine images is expected to benefit clinical decision support systems for diagnosis, surgery planning, and population-based analysis of spine and bone health. However, designing automated algorithms for spine processing is challenging predominantly due to considerable variations in anatomy and acquisition protocols and due to a severe shortage of publicly available data. Addressing these limitations, the Large Scale Vertebrae Segmentation Challenge (VerSe) was organised in conjunction with the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) in 2019 and 2020, with a call for algorithms tackling the labelling and segmentation of vertebrae. Two datasets containing a total of 374 multi-detector CT scans from 355 patients were prepared and 4505 vertebrae have individually been annotated at voxel level by a human-machine hybrid algorithm (https://osf.io/nqjyw/, https://osf.io/t98fz/). A total of 25 algorithms were benchmarked on these datasets. In this work, we present the results of this evaluation and further investigate the performance variation at the vertebra level, scan level, and different fields of view. We also evaluate the generalisability of the approaches to an implicit domain shift in data by evaluating the top-performing algorithms of one challenge iteration on data from the other iteration. The principal takeaway from VerSe: the performance of an algorithm in labelling and segmenting a spine scan hinges on its ability to correctly identify vertebrae in cases of rare anatomical variations. The VerSe content and code can be accessed at: https://github.com/anjany/verse.

Clinical implementation of automated treatment planning for whole-brain radiotherapy.

The purpose of the study was to develop and clinically deploy an automated, deep learning-based approach to treatment planning for whole-brain radiotherapy (WBRT). We collected CT images and radiotherapy treatment plans to automate a beam aperture definition from 520 patients who received WBRT. These patients were split into training (n = 312), cross-validation (n = 104), and test (n = 104) sets which were used to train and evaluate a deep learning model. The DeepLabV3+ architecture was trained to automatically define the beam apertures on lateral-opposed fields using digitally reconstructed radiographs (DRRs). For the beam aperture evaluation, 1st quantitative analysis was completed using a test set before clinical deployment and 2nd quantitative analysis was conducted 90 days after clinical deployment. The mean surface distance and the Hausdorff distances were compared in the anterior-inferior edge between the clinically used and the predicted fields. Clinically used plans and deep-learning generated plans were evaluated by various dose-volume histogram metrics of brain, cribriform plate, and lens. The 1st quantitative analysis showed that the average mean surface distance and Hausdorff distance were 7.1 mm (±3.8 mm) and 11.2 mm (±5.2 mm), respectively, in the anterior-inferior edge of the field. The retrospective dosimetric comparison showed that brain dose coverage (D99%, D95%, D1%) of the automatically generated plans was 29.7, 30.3, and 32.5 Gy, respectively, and the average dose of both lenses was up to 19.0% lower when compared to the clinically used plans. Following the clinical deployment, the 2nd quantitative analysis showed that the average mean surface distance and Hausdorff distance between the predicted and clinically used fields were 2.6 mm (±3.2 mm) and 4.5 mm (±5.6 mm), respectively. In conclusion, the automated patient-specific treatment planning solution for WBRT was implemented in our clinic. The predicted fields appeared consistent with clinically used fields and the predicted plans were dosimetrically comparable.

Effects of alterations in positron emission tomography imaging parameters on radiomics features.

Radiomics studies require large patient cohorts, which often include patients imaged using different imaging protocols. We aimed to determine the impact of variability in imaging protocol parameters and interscanner variability using a phantom that produced feature values similar to those of patients. Positron emission tomography (PET) scans of a Hoffman brain phantom were acquired on GE Discovery 710, Siemens mCT, and Philips Vereos scanners. A standard-protocol scan was acquired on each machine, and then each parameter that could be changed was altered individually. The phantom was contoured with 10 regions of interest (ROIs). Values for 45 features with 2 different preprocessing techniques were extracted for each image. To determine the impact of each parameter on the reliability of each radiomics feature, the intraclass correlation coefficient (ICC) was calculated with the ROIs as the subjects and the parameter values as the raters. For interscanner comparisons, we compared the standard deviation of each radiomics feature value from the standard-protocol images to the standard deviation of the same radiomics feature from PET scans of 224 patients with non-small cell lung cancer. When the pixel size was resampled prior to feature extraction, all features had good reliability (ICC > 0.75) for the field of view and matrix size. The time per bed position had excellent reliability (ICC > 0.9) on all features. When the filter cutoff was restricted to values below 6 mm, all features had good reliability. Similarly, when subsets and iterations were restricted to reasonable values used in clinics, almost all features had good reliability. The average ratio of the standard deviation of features on the phantom scans to that of the NSCLC patient scans was 0.73 using fixed-bin-width preprocessing and 0.92 using 64-level preprocessing. Most radiomics feature values had at least good reliability when imaging protocol parameters were within clinically used ranges. However, interscanner variability was about equal to interpatient variability; therefore, caution must be used when combining patients scanned on equipment from different vendors in radiomics data sets.