Machine learning automated treatment planning for online magnetic resonance guided adaptive radiotherapy of prostate cancer.

Background and purpose: No best practices currently exist for achieving high quality radiation therapy (RT) treatment plan adaptation during magnetic resonance (MR) guided RT of prostate cancer. This study validates the use of machine learning (ML) automated RT treatment plan adaptation and benchmarks it against current clinical RT plan adaptation methods. Materials and methods: We trained an atlas-based ML automated treatment planning model using reference MR RT treatment plans (42.7 Gy in 7 fractions) from 46 patients with prostate cancer previously treated at our institution. For a held-out test set of 38 patients, retrospectively generated ML RT plans were compared to clinical human-generated adaptive RT plans for all 266 fractions. Differences in dose-volume metrics and clinical objective pass rates were evaluated using Wilcoxon tests (p < 0.05) and Exact McNemar tests (p < 0.05), respectively. Results: Compared to clinical RT plans, ML RT plans significantly increased sparing and objective pass rates of the rectum, bladder, and left femur. The mean ± standard deviation of rectum D20 and D50 in ML RT plans were 2.5 ± 2.2 Gy and 1.6 ± 1.3 Gy lower than clinical RT plans, respectively, with 14 % higher pass rates; bladder D40 was 4.6 ± 2.9 Gy lower with a 20 % higher pass rate; and the left femur D5 was 0.8 ± 1.8 Gy lower with a 7 % higher pass rate. Conclusions: ML automated RT treatment plan adaptation increases robustness to interfractional anatomical changes compared to current clinical adaptive RT practices by increasing compliance to treatment objectives.

A T2-weighted MRI-based radiomic signature for disease-free survival in locally advanced cervical cancer following chemoradiation: An international, multicentre study.

INTRODUCTION: To develop and validate a T2-weighted magnetic resonance imaging (MRI)-based radiomic signature associated with disease-free survival (DFS) in locally advanced cervical cancer. MATERIALS AND METHODS: The study comprised a training dataset of 132 patients (93 Norwegian; 39 The Cancer Imaging Archive (TCIA) and an independent validation Canadian dataset of 199 patients with FIGO stage IB-IVA cervical cancer treated with chemoradiation. Radiomic features were extracted using PyRadiomics. A radiomic signature was developed based on a multivariable radiomic prognostic model for DFS built using the training dataset, with minimal redundancy maximum relevancy feature selection method. Univariate and multivariable Cox regression analyses were then conducted to examine the association of the derived radiomic signature with DFS. RESULTS: A radiomic signature was prognostic for DFS in the training cohort (Norwegian hazard ratio [HR] 5.54, p = 0.002; TCIA HR 3.59, p = 0.04). The radiomic signature remained independently associated with DFS (HR 3.70, p = 0.004) when adjusted for stage and tumor volume. The radiomic signature was also prognostic for DFS in the validation cohort, both on univariate analysis (HR 2.22, p = 0.003), and multivariable analysis adjusted for stage and tumor volume (HR 1.84, p = 0.04). The 4-year DFS rates of patients with radiomic signature score > 0 vs ≤ 0 were 48.2 % vs 87.9 %, and 56.4 % vs 80.8 % for training and validation cohorts respectively. CONCLUSION: An MRI-based radiomic signature can be used as a prognostic biomarker for DFS in patients with locally advanced cervical cancer undergoing chemoradiation.

Shortcut learning in medical AI hinders generalization: method for estimating AI model generalization without external data.

Healthcare datasets are becoming larger and more complex, necessitating the development of accurate and generalizable AI models for medical applications. Unstructured datasets, including medical imaging, electrocardiograms, and natural language data, are gaining attention with advancements in deep convolutional neural networks and large language models. However, estimating the generalizability of these models to new healthcare settings without extensive validation on external data remains challenging. In experiments across 13 datasets including X-rays, CTs, ECGs, clinical discharge summaries, and lung auscultation data, our results demonstrate that model performance is frequently overestimated by up to 20% on average due to shortcut learning of hidden data acquisition biases (DAB). Shortcut learning refers to a phenomenon in which an AI model learns to solve a task based on spurious correlations present in the data as opposed to features directly related to the task itself. We propose an open source, bias-corrected external accuracy estimate, PEst, that better estimates external accuracy to within 4% on average by measuring and calibrating for DAB-induced shortcut learning.

Repeatability and reproducibility of prostate apparent diffusion coefficient values on a 1.5 T magnetic resonance linear accelerator.

Background and Purpose: Integrated magnetic resonance linear accelerator (MR-Linac) systems offer potential for biologically based adaptive radiation therapy using apparent diffusion coefficient (ADC). Accurate tracking of longitudinal ADC changes is key to establishing ADC-driven dose adaptation. Here, we report repeatability and reproducibility of intraprostatic ADC using deformable image registration (DIR) to correct for inter-fraction prostate changes. Materials and Methods: The study included within-fraction repeat ADC measurements for three consecutive fractions for 20 patients with prostate cancer treated on a 1.5 T MR-Linac. We deformably registered successive fraction T2-weighted images and applied the deformation vector field to corresponding ADC maps to align to fraction 2. We delineated gross tumour volume (GTV), peripheral zone (PZ) and prostate clinical target volume (CTV) regions-of-interest (ROIs) on T2-weighted MRI and copied to ADC maps. We computed intraclass correlation coefficients (ICC) and percent repeatability coefficient (%RC) to determine within-fraction repeatability and between-fraction reproducibility for individual voxels, mean and 10th percentile ADC values per ROI. Results: The ICC between repeats and fractions was excellent for mean and 10th percentile ADC in all ROIs (ICC > 0.86), and moderate repeatability and reproducibility existed for individual voxels (ICC > 0.542). Similarly, low %RC within-fraction (4.2-17.9 %) mean and 10th percentile ADC existed, with greater %RC between fractions (10.2-36.8 %). Higher %RC existed for individual voxel within-fraction (21.7-30.6 %) and between-fraction (32.1-34.5 %) ADC. Conclusions: Results suggest excellent ADC repeatability and reproducibility in clinically relevant ROIs using DIR to correct between-fraction anatomical changes. We established the precision of voxel-level ADC tracking for future biologically based adaptation implementation.

Expanding Access to Computed Tomographic Staging and Three-Dimensional Intensity Modulated Radiotherapy for Cervical Cancer in Ghana.

PURPOSE: To build capacity for improved treatment of locally advanced cervical cancer in Ghana, including computed tomography (CT) staging and intensity modulated radiotherapy (IMRT). MATERIALS AND METHODS: Patients with histologically confirmed cervical cancer were prospectively staged with abdominopelvic CT and ultrasound and offered the opportunity to have IMRT instead of conventional two-dimensional radiotherapy. The development of an efficient, high-quality, and safe IMRT program was facilitated by investment in new technology and comprehensive training of the interdisciplinary radiotherapy team in collaboration with a North American center of excellence. RESULTS: Of 215 patients with cervical cancer referred in 2022, 66% were able to afford CT scans and 26% were able to afford IMRT. Lymph node metastases were identified in 52% of patients by CT but in only 2% of patients by ultrasound. The use of CT resulted in 63% of patients being upstaged and changed treatment intent or radiation treatment volumes in 67% of patients. Patients who had IMRT experienced fewer acute side effects and were more likely to complete treatment as planned. CONCLUSION: It is feasible to provide state-of the-art cancer treatment with CT staging and IMRT to patients with cervical cancer in low-resource settings and achieve meaningful improvements in outcomes. It requires a broad commitment by program leadership to invest in technology and staff training. Major challenges include balancing improved clinical care with reduced patient throughput when radiation treatment capacity is constrained, and with the additional cost in the absence of universal health coverage.

RADCURE: An open-source head and neck cancer CT dataset for clinical radiation therapy insights.

PURPOSE: This manuscript presents RADCURE, one of the most extensive head and neck cancer (HNC) imaging datasets accessible to the public. Initially collected for clinical radiation therapy (RT) treatment planning, this dataset has been retrospectively reconstructed for use in imaging research. ACQUISITION AND VALIDATION METHODS: RADCURE encompasses data from 3346 patients, featuring computed tomography (CT) RT simulation images with corresponding target and organ-at-risk contours. These CT scans were collected using systems from three different manufacturers. Standard clinical imaging protocols were followed, and contours were manually generated and reviewed at weekly RT quality assurance rounds. RADCURE imaging and structure set data was extracted from our institution's radiation treatment planning and oncology information systems using a custom-built data mining and processing system. Furthermore, images were linked to our clinical anthology of outcomes data for each patient and includes demographic, clinical and treatment information based on the 7th edition TNM staging system (Tumor-Node-Metastasis Classification System of Malignant Tumors). The median patient age is 63, with the final dataset including 80% males. Half of the cohort is diagnosed with oropharyngeal cancer, while laryngeal, nasopharyngeal, and hypopharyngeal cancers account for 25%, 12%, and 5% of cases, respectively. The median duration of follow-up is five years, with 60% of the cohort surviving until the last follow-up point. DATA FORMAT AND USAGE NOTES: The dataset provides images and contours in DICOM CT and RT-STRUCT formats, respectively. We have standardized the nomenclature for individual contours-such as the gross primary tumor, gross nodal volumes, and 19 organs-at-risk-to enhance the RT-STRUCT files' utility. Accompanying demographic, clinical, and treatment data are supplied in a comma-separated values (CSV) file format. This comprehensive dataset is publicly accessible via The Cancer Imaging Archive. POTENTIAL APPLICATIONS: RADCURE's amalgamation of imaging, clinical, demographic, and treatment data renders it an invaluable resource for a broad spectrum of radiomics image analysis research endeavors. Researchers can utilize this dataset to advance routine clinical procedures using machine learning or artificial intelligence, to identify new non-invasive biomarkers, or to forge prognostic models.

Clinical and dosimetric outcomes of image-guided, dose-painted radiotherapy in muscle invasive bladder cancer.

PURPOSE/OBJECTIVE: Definitive radiotherapy (RT) is an alternative to radical cystectomy for select patients with muscle invasive bladder cancer (MIBC); however, there is limited data on dose-painted RT approaches. We report the clinical and dosimetric outcomes of a cohort of MIBC patients treated with dose-painted RT. MATERIAL/METHODS: This was a single institution retrospective study of cT2-4N0M0 MIBC patients treated with external beam radiotherapy (EBRT) to the bladder, and sequential or concomitant boost to the tumor bed. The target delineation was guided by either intravesical injection of Lipiodol or through fusion of the pre-treatment imaging. The majority were treated with daily image-guidance. Kaplan-Meier was used to characterize overall survival (OS) and progression-free survival (PFS). Cumulative incidence function (CIF) was used to estimate local (intravesical) recurrence (LR), regional recurrence (RR) and distant metastasis (DM). Univariable and multivariable cause-specific hazard model was used to assess factors associated with LR and OS. RESULTS: 117 patients were analyzed. The median age was 73 years (range 43, 95). The median EQD2 to the boost volume was 66 Gy (range 52.1, 70). Lipiodol injection was used in 64 patients (55%), all treated with IMRT/VMAT. 95 (81%) received concurrent chemotherapy, of whom, 44 (38%) received neoadjuvant chemotherapy. The median follow-up was 37 months (IQR 16.2, 83.3). At 5-year, OS and PFS were 79% (95% CI 70.5-89.2) and 46% (95% CI 36.5-57.5). Forty-five patients had bladder relapse, of which 30 patients (67%) were at site of the tumor bed. Nine patients underwent salvage-cystectomy. Late high-grade (G3-G4) genitourinary and gastrointestinal toxicity were 3% and 1%. CONCLUSION: Partial boost RT in MIBC is associated with good local disease control and high rates of cystectomy free survival. We observed a pattern of predominantly LR in the tumor bed, supporting the use of a dose-painted approach/de-escalation strategy to the uninvolved bladder. Prospective trials are required to compare oncological and toxicity outcomes between dose-painted and homogeneous bladder RT techniques.

A total inverse planning paradigm: Prospective clinical trial evaluating the performance of a novel MR-based 3D-printed head immobilization device.

Background and purpose: Brain radiotherapy (cnsRT) requires reproducible positioning and immobilization, attained through redundant dedicated imaging studies and a bespoke moulding session to create a thermoplastic mask (T-mask). Innovative approaches may improve the value of care. We prospectively deployed and assessed the performance of a patient-specific 3D-printed mask (3Dp-mask), generated solely from MR imaging, to replicate a reproducible positioning and tolerable immobilization for patients undergoing cnsRT. Material and methods: Patients undergoing LINAC-based cnsRT (primary tumors or resected metastases) were enrolled into two arms: control (T-mask) and investigational (3Dp-mask). For the latter, an in-house designed 3Dp-mask was generated from MR images to recreate the head positioning during MR acquisition and allow coupling with the LINAC tabletop. Differences in inter-fraction motion were compared between both arms. Tolerability was assessed using patient-reported questionnaires at various time points. Results: Between January 2020 - July 2022, forty patients were enrolled (20 per arm). All participants completed the prescribed cnsRT and study evaluations. Average 3Dp-mask design and printing completion time was 36 h:50 min (range 12 h:56 min - 42 h:01 min). Inter-fraction motion analyses showed three-axis displacements comparable to the acceptable tolerance for the current standard-of-care. No differences in patient-reported tolerability were seen at baseline. During the last week of cnsRT, 3Dp-mask resulted in significantly lower facial and cervical discomfort and patients subjectively reported less pressure and confinement sensation when compared to the T-mask. No adverse events were observed. Conclusion: The proposed total inverse planning paradigm using a 3D-printed immobilization device is feasible and renders comparable inter-fraction performance while offering a better patient experience, potentially improving cnsRT workflows and its cost-effectiveness.

Multi-institutional Prognostic Modeling in Head and Neck Cancer: Evaluating Impact and Generalizability of Deep Learning and Radiomics.

Artificial intelligence (AI) and machine learning (ML) are becoming critical in developing and deploying personalized medicine and targeted clinical trials. Recent advances in ML have enabled the integration of wider ranges of data including both medical records and imaging (radiomics). However, the development of prognostic models is complex as no modeling strategy is universally superior to others and validation of developed models requires large and diverse datasets to demonstrate that prognostic models developed (regardless of method) from one dataset are applicable to other datasets both internally and externally. Using a retrospective dataset of 2,552 patients from a single institution and a strict evaluation framework that included external validation on three external patient cohorts (873 patients), we crowdsourced the development of ML models to predict overall survival in head and neck cancer (HNC) using electronic medical records (EMR) and pretreatment radiological images. To assess the relative contributions of radiomics in predicting HNC prognosis, we compared 12 different models using imaging and/or EMR data. The model with the highest accuracy used multitask learning on clinical data and tumor volume, achieving high prognostic accuracy for 2-year and lifetime survival prediction, outperforming models relying on clinical data only, engineered radiomics, or complex deep neural network architecture. However, when we attempted to extend the best performing models from this large training dataset to other institutions, we observed significant reductions in the performance of the model in those datasets, highlighting the importance of detailed population-based reporting for AI/ML model utility and stronger validation frameworks. We have developed highly prognostic models for overall survival in HNC using EMRs and pretreatment radiological images based on a large, retrospective dataset of 2,552 patients from our institution.Diverse ML approaches were used by independent investigators. The model with the highest accuracy used multitask learning on clinical data and tumor volume.External validation of the top three performing models on three datasets (873 patients) with significant differences in the distributions of clinical and demographic variables demonstrated significant decreases in model performance. Significance: ML combined with simple prognostic factors outperformed multiple advanced CT radiomics and deep learning methods. ML models provided diverse solutions for prognosis of patients with HNC but their prognostic value is affected by differences in patient populations and require extensive validation.

Simple and effective immobilization for radiation treatment of choroidal melanoma.

At our institution, patients diagnosed with choroidal melanoma requiring external beam radiation therapy are treated with two 6 MV volumetric-modulated arcs delivering 50 Gy over 5 daily fractions. The patient is immobilized using an Orfit head and neck mask and is directed to look at a light emitting diode (LED) during CT simulation and treatment to minimize eye movement. Patient positioning is checked with cone beam computed tomography (CBCT) daily. Translational and rotational displacements greater than 1 mm or 1° off the planned isocenter position are corrected using a Hexapod couch. The aim of this study is to verify that the mask system provides adequate immobilization and to verify our 2-mm planning target volume (PTV) margins are sufficient. Residual displacements provided by pretreatment verification and post-treatment CBCT data sets were used to assess the impact of patient mobility during treatment on the reconstructed delivered dose to the target and organs at risk. The PTV margin calculated using van Herk's method1 was used to assess patient motion plus other factors that affect treatment position, such as kV-MV isocenter coincidence. Patient position variations were small and were shown to not cause significant dose variations between the planned and reconstructed dose to the target and organs at risk. The PTV margin analysis showed patient translational motion alone required a PTV margin of 1 mm. Given other factors that affect treatment delivery accuracy, a 2-mm PTV margin was shown to be sufficient for treatment of 95% of our patients with 100% of dose delivered to the GTV. The mask immobilization with LED focus is robust and we showed a 2-mm PTV margin is adequate with this technique.

Liver SBRT dose accumulation to assess the impact of anatomic variations on normal tissue doses and toxicity in patients treated with concurrent sorafenib.

BACKGROUND AND PURPOSE: Unexpected liver volume reductions occurred during trials of liver SBRT and concurrent sorafenib. The aims were to accumulate liver SBRT doses to assess the impact of these anatomic variations on normal tissue dose parameters and toxicity. MATERIALS AND METHODS: Thirty-two patients with hepatocellular carcinoma (HCC) or metastases treated on trials of liver SBRT (30-57 Gy, 6 fractions) and concurrent sorafenib were analyzed. SBRT doses were accumulated using biomechanical deformable registration of daily cone-beam CT. Dose deviations (accumulated-planned) for normal tissues were compared for patients with liver volume reductions > 100 cc versus stable volumes, and accumulated doses were reported for three patients with grade 3-5 luminal gastrointestinal toxicities. RESULTS: Patients with reduced (N = 12) liver volumes had larger mean deviations of 0.4-1.3 Gy in normal tissues, versus -0.2-0.4 Gy for stable cases (N = 20), P > 0.05. Deviations > 5% of the prescribed dose occurred in both groups. Two HCC patients with toxicities to small and large bowel had liver volume reductions and deviations to the maximum dose of 4% (accumulated 36.9 Gy) and 3% (accumulated 33.4 Gy) to these organs respectively. Another HCC patient with a toxicity of unknown location plus tumor rupture, had stable liver volumes and deviations to luminal organs of -6% to 4.5% (accumulated < 30.5 Gy). CONCLUSION: Liver volume reductions during SBRT and concurrent sorafenib were associated with larger increases in accumulated dose to normal tissues versus stable liver volumes. These dosimetric changes may have further contributed to toxicities in HCC patients who have higher baseline risks.

Evaluation of dosimetric predictors of toxicity after IMRT with concurrent chemotherapy for anal cancer.

BACKGROUND: This study investigates the impact of dosimetric parameters on acute and late toxicity for patients with anal squamous cell carcinoma (SCC) treated with image-guided intensity modulated radiation therapy (IG-IMRT) and concurrent chemotherapy. MATERIALS AND METHODS: Patients were enrolled in an observational cohort study between 2008 and 2013 (median follow-up 3.4 years). They were treated with standardized target and organ-at-risk (OAR) contouring, planning, and IG-IMRT. Radiotherapy dose, based on clinicopathologic features, ranged from 45 Gy to 63 Gy to gross targets and 27 Gy to 36 Gy to elective targets. Chemotherapy was concurrent 5-fluorouracil and mitomycin C (weeks 1&5). Toxicity was prospectively graded using NCI CTCAE v.3 and RTOG scales. Logistic regression was used to assess the association between dose/volume parameters (e.g small bowel V5) and corresponding grade 2 + and 3+ (G2+/3 + ) toxicities (e.g. diarrhea). RESULTS: In total, 87 and 79 patients were included in the acute and late toxicity analyses, respectively. The most common acute G2 + toxicities were skin (dermatitis in 87 % [inguino-genital skin], 91 % [perianal skin]) and hematologic in 58 %. G2 + late anal toxicity (sphincter dysfunction), gastrointestinal toxicity, and skin toxicity were respectively experienced by 49 %, 38 %, and 44 % of patients. Statistically significant associations were observed between: G2 + acute diarrhea and small bowel V35; G2 + acute genitourinary toxicity and bladder D0.5cc; G2 + inguino-genital skin toxicity and anterior skin V35; G2 + perianal skin toxicity and posterior skin V15; G2 + anemia and lower pelvis bone V45. D0.5 cc was significantly predictive of late toxicity (G2 + anal dysfunction, intestinal toxicity, and inguino-genital/perianal dermatitis). Maximum skin toxicity grade was significantly correlated with the requirement for a treatment break. CONCLUSION: Statistically significant dose-volume parameters were identified and may be used to offer individualized risk prediction and to inform treatment planning. Additional validation of the results is required.

Time-course assessment of 3D-image distortion on the 1.5 T Marlin/Elekta Unity MR-LINAC.

PURPOSE: The efficacy of MR-guided radiotherapy on a MR-LINAC (MR-L) is dependent on the geometric accuracy of its MR images over clinically relevant Fields-of-View (FOVs). Our objectives were to: evaluate gradient non-linearity (GNL) on the Elekta Unity MR-L across time via 76 weekly measurements of 3D-distortion over concentrically larger diameter spherical volumes (DSVs); quantify distortion measurement error; and assess the temporal stability of spatial distortion using statistical process control (SPC). METHODS: MR-image distortion was assessed using a large-FOV 3D-phantom containing 1932 markers embedded in seven parallel plates, spaced 25 mm × 25 mm in- and 55 mm through-plane. Automatically analyzed T1 images yielded distortions in 200, 300, 400 and 500 mm concentric DSVs. Distortion measurement error was evaluated using median absolute difference analysis of imaging repeatability tests. RESULTS: Over the measurement period absolute time-averaged distortion varied between: dr = 0.30 - 0.49 mm, 0.53 - 0.80 mm, 1.0 - 1.4 mm and 2.28 - 2.37 mm, for DSVs 200, 300, 400 and 500 mm at the 98th percentile level. Repeatability tests showed that imaging/repositioning introduces negligible error: mean ≤ 0.02 mm (max ≤ 0.3 mm). SPC analysis showed image distortion was stable across all DSVs; however, noticeable changes in GNL were observed following servicing at the one-year mark. CONCLUSIONS: Image distortion on the MR-L is in the sub-millimeter range for DSVs ≤ 300 mm and stable across time, with SPC analysis indicating all measurements remain within control for each DSV.

Dose accumulation for MR-guided adaptive radiotherapy: From practical considerations to state-of-the-art clinical implementation.

MRI-linear accelerator (MR-linac) devices have been introduced into clinical practice in recent years and have enabled MR-guided adaptive radiation therapy (MRgART). However, by accounting for anatomical changes throughout radiation therapy (RT) and delivering different treatment plans at each fraction, adaptive radiation therapy (ART) highlights several challenges in terms of calculating the total delivered dose. Dose accumulation strategies-which typically involve deformable image registration between planning images, deformable dose mapping, and voxel-wise dose summation-can be employed for ART to estimate the delivered dose. In MRgART, plan adaptation on MRI instead of CT necessitates additional considerations in the dose accumulation process because MRI pixel values do not contain the quantitative information used for dose calculation. In this review, we discuss considerations for dose accumulation specific to MRgART and in relation to current MR-linac clinical workflows. We present a general dose accumulation framework for MRgART and discuss relevant quality assurance criteria. Finally, we highlight the clinical importance of dose accumulation in the ART era as well as the possible ways in which dose accumulation can transform clinical practice and improve our ability to deliver personalized RT.

Simulated dose painting of hypoxic sub-volumes in pancreatic cancer stereotactic body radiotherapy.

Dose painting of hypoxic tumour sub-volumes using positron-emission tomography (PET) has been shown to improve tumour controlin silicoin several sites, predominantly head and neck and lung cancers. Pancreatic cancer presents a more stringent challenge, given its proximity to critical gastro-intestinal organs-at-risk (OARs), anatomic motion, and impediments to reliable PET hypoxia quantification. A radiobiological model was developed to estimate clonogen survival fraction (SF), using18F-fluoroazomycin arabinoside PET (FAZA PET) images from ten patients with unresectable pancreatic ductal adenocarcinoma to quantify oxygen enhancement effects. For each patient, four simulated five-fraction stereotactic body radiotherapy (SBRT) plans were generated: (1) a standard SBRT plan aiming to cover the planning target volume with 40 Gy, (2) dose painting plans delivering escalated doses to a maximum of three FAZA-avid hypoxic sub-volumes, (3) dose painting plans with simulated spacer separating the duodenum and pancreatic head, and (4), plans with integrated boosts to geometric contractions of the gross tumour volume (GTV). All plans saturated at least one OAR dose limit. SF was calculated for each plan and sensitivity of SF to simulated hypoxia quantification errors was evaluated. Dose painting resulted in a 55% reduction in SF as compared to standard SBRT; 78% with spacer. Integrated boosts to hypoxia-blind geometric contractions resulted in a 41% reduction in SF. The reduction in SF for dose-painting plans persisted for all hypoxia quantification parameters studied, including registration and rigid motion errors that resulted in shifts and rotations of the GTV and hypoxic sub-volumes by as much as 1 cm and 10 degrees. Although proximity to OARs ultimately limited dose escalation, with estimated SFs (∼10-5) well above levels required to completely ablate a ∼10 cm3tumour, dose painting robustly reduced clonogen survival when accounting for expected treatment and imaging uncertainties and thus, may improve local response and associated morbidity.

Limited-stage small cell lung cancer: Outcomes associated with prophylactic cranial irradiation over a 20-year period at the Princess Margaret Cancer Centre.

BACKGROUND & PURPOSE: Prophylactic cranial irradiation (PCI) is recommended for limited-stage small-cell lung cancer (LS-SCLC) patients with good response to concurrent chemoradiation. We report our institution's 20-year experience with this patient population and associated clinical outcomes. MATERIALS & METHODS: A retrospective cohort of consecutive LS-SCLC patients treated with curative intent chemoradiation at our institution (1997-2018) was reviewed. Overall survival (OS) was calculated using the Kaplan-Meier method, and significant covariates determined by the Cox proportional hazards model. Covariates predictive of PCI were determined using Fisher's exact test and the Mann-Whitney test. Brain failure risk (BFR) was calculated using the cumulative incidence method treating death as a competing event. Treatment cohorts (historic vs. contemporary) were stratified by the median year of diagnosis (2005). RESULTS: A total of 369 patients with LS-SCLC were identified, of which 278 patients were notionally PCI eligible. PCI was given to 196 patients (71%). Younger age was associated with PCI utilization (p < 0.001). PCI utilization rates did not change between the historic and contemporary treatment era (p = 0.11), whereas magnetic resonance imaging (MRI) use at baseline and follow-up became more prevalent in the contemporary era (p = <0.001). On multivariable analysis, PCI utilization was associated with improved OS (HR 1.88, 95% CI 1.32-2.69) and decreased BFR (HR 4.66, 95% CI 2.58-8.40). Patients who had MRI follow-up had a higher incidence of BFR (HR 0.35, 95% CI 0.18-0.66) in multivariable analyses. CONCLUSIONS: For LS-SCLC patients at our institution, PCI is more frequently utilized in younger patients, and the utilization rate did not change significantly over the past 20 years. PCI was independently associated with improved OS and lower BFR. Omission of PCI in LS-SCLC patients should not be routinely practiced in the absence of further prospective data.

Automated detection of dental artifacts for large-scale radiomic analysis in radiation oncology.

BACKGROUND AND PURPOSE: Computed tomography (CT) is one of the most common medical imaging modalities in radiation oncology and radiomics research, the computational voxel-level analysis of medical images. Radiomics is vulnerable to the effects of dental artifacts (DA) caused by metal implants or fillings and can hamper future reproducibility on new datasets. In this study we seek to better understand the robustness of quantitative radiomic features to DAs. Furthermore, we propose a novel method of detecting DAs in order to safeguard radiomic studies and improve reproducibility. MATERIALS AND METHODS: We analyzed the correlations between radiomic features and the location of dental artifacts in a new dataset containing 3D CT scans from 3211 patients. We then combined conventional image processing techniques with a pre-trained convolutional neural network to create a three-class patient-level DA classifier and slice-level DA locator. Finally, we demonstrated its utility in reducing the correlations between the location of DAs and certain radiomic features. RESULTS: We found that when strong DAs were present, the proximity of the tumour to the mouth was highly correlated with 36 radiomic features. We predicted the correct DA magnitude yielding a Matthews correlation coefficient of 0.73 and location of DAs achieving the same level of agreement as human labellers. CONCLUSIONS: Removing radiomic features or CT slices containing DAs could reduce the unwanted correlations between the location of DAs and radiomic features. Automated DA detection can be used to improve the reproducibility of radiomic studies; an important step towards creating effective radiomic models for use in clinical radiation oncology.

Prediction of Human Papillomavirus (HPV) Association of Oropharyngeal Cancer (OPC) Using Radiomics: The Impact of the Variation of CT Scanner.

Studies have shown that radiomic features are sensitive to the variability of imaging parameters (e.g., scanner models), and one of the major challenges in these studies lies in improving the robustness of quantitative features against the variations in imaging datasets from multi-center studies. Here, we assess the impact of scanner choice on computed tomography (CT)-derived radiomic features to predict the association of oropharyngeal squamous cell carcinoma with human papillomavirus (HPV). This experiment was performed on CT image datasets acquired from two different scanner manufacturers. We demonstrate strong scanner dependency by developing a machine learning model to classify HPV status from radiological images. These experiments reveal the effect of scanner manufacturer on the robustness of radiomic features, and the extent of this dependency is reflected in the performance of HPV prediction models. The results of this study highlight the importance of implementing an appropriate approach to reducing the impact of imaging parameters on radiomic features and consequently on the machine learning models, without removing features which are deemed non-robust but may contain learning information.

The impact of the variation of imaging parameters on the robustness of Computed Tomography radiomic features: A review.

The field of radiomics is at the forefront of personalized medicine. However, there is concern that high variation in imaging parameters will impact robustness of radiomic features and subsequently the performance of the predictive models built upon them. Therefore, our review aims to evaluate the impact of imaging parameters on the robustness of radiomic features. We also provide insights into the validity and discrepancy of different methodologies applied to investigate the robustness of radiomic features. We selected 47 papers based on our predefined inclusion criteria and grouped these papers by the imaging parameter under investigation: (i) scanner parameters, (ii) acquisition parameters and (iii) reconstruction parameters. Our review highlighted that most of the imaging parameters are disruptive parameters, and shape along with First order statistics were reported as the most robust radiomic features against variation in imaging parameters. This review identified inconsistencies related to the methodology of the reviewed studies such as the metrics used for robustness, the feature extraction techniques, the reporting style, and their outcome inclusion. We hope this review will aid the scientific community in conducting research in a way that is more reproducible and avoids the pitfalls of previous analyses.

Practical Considerations for the Implementation of a Stereotactic Body Radiation Therapy Program for Oligo-Metastases.

PURPOSE: With multiple phase 2 trials supporting the use of stereotactic body radiation therapy (SBRT) in oligo-metastatic disease, we evaluated practices that could inform effective implementation of an oligo-metastasis SBRT program. METHODS AND MATERIALS: Using a context-focused realist methodology, an advisory committee of interprofessional clinicians met over a series of semistructured teleconference meetings to identify challenges in implementing an oligo-metastasis SBRT program. Consideration was given to 2 models of care: a subspecialist anatomic expertise model versus a single-practitioner "quarterback" model. RESULTS: The advisory committee structured recommendations within a context-mechanism-outcome framework. In summary, the committee recommends that during patient workup, a single practitioner arranges the minimum number of necessary tests, with case presentation at an appropriate multidisciplinary tumor board, including careful review of all previous treatments, and enrollment on clinical trials when possible. At simulation, common patient positions and immobilization on a single simulation scan for multiple sites is recommended. During radiation planning, dose-fractionation regimens should safely facilitate cumulative dose calculations, a single isocenter should be considered for multiple close targets to reduce treatment time, and adherence to strict quality assurance protocols is strongly recommended. Treatment duration should be minimized by treating multiple sites on the same day or choosing shorter dose fractionations. Team communication, thorough documentation, and standardized nomenclature can reduce system errors. Follow-up should aim to minimize redundant clinical appointments and imaging scans. Expert radiology review may be required to interpret post-SBRT imaging. CONCLUSIONS: These guidelines inform best clinical practices for implementing an oligo-metastasis SBRT program. Iterations using a realist approach may further expand on local contexts.