Consistency in contouring of organs at risk by artificial intelligence vs oncologists in head and neck cancer patients.

BACKGROUND: In the Danish Head and Neck Cancer Group (DAHANCA) 35 trial, patients are selected for proton treatment based on simulated reductions of Normal Tissue Complication Probability (NTCP) for proton compared to photon treatment at the referring departments. After inclusion in the trial, immobilization, scanning, contouring and planning are repeated at the national proton centre. The new contours could result in reduced expected NTCP gain of the proton plan, resulting in a loss of validity in the selection process. The present study evaluates if contour consistency can be improved by having access to AI (Artificial Intelligence) based contours. MATERIALS AND METHODS: The 63 patients in the DAHANCA 35 pilot trial had a CT from the local DAHANCA centre and one from the proton centre. A nationally validated convolutional neural network, based on nnU-Net, was used to contour OARs on both scans for each patient. Using deformable image registration, local AI and oncologist contours were transferred to the proton centre scans for comparison. Consistency was calculated with the Dice Similarity Coefficient (DSC) and Mean Surface Distance (MSD), comparing contours from AI to AI and oncologist to oncologist, respectively. Two NTCP models were applied to calculate NTCP for xerostomia and dysphagia. RESULTS: The AI contours showed significantly better consistency than the contours by oncologists. The median and interquartile range of DSC was 0.85 [0.78 - 0.90] and 0.68 [0.51 - 0.80] for AI and oncologist contours, respectively. The median and interquartile range of MSD was 0.9 mm [0.7 - 1.1] mm and 1.9 mm [1.5 - 2.6] mm for AI and oncologist contours, respectively. There was no significant difference in ΔNTCP. CONCLUSIONS: The study showed that OAR contours made by the AI algorithm were more consistent than those made by oncologists. No significant impact on the ΔNTCP calculations could be discerned.

An open source auto-segmentation algorithm for delineating heart and substructures - Development and validation within a multicenter lung cancer cohort.

BACKGROUND AND PURPOSE: Irradiation of the heart in thoracic cancers raises toxicity concerns. For accurate dose estimation, automated heart and substructure segmentation is potentially useful. In this study, a hybrid automatic segmentation is developed. The accuracy of delineation and dose predictions were evaluated, testing the method's potential within heart toxicity studies. MATERIALS AND METHODS: The hybrid segmentation method delineated the heart, four chambers, three large vessels, and the coronary arteries. The method consisted of a nnU-net heart segmentation and partly atlas- and model-based segmentation of the substructures. The nnU-net training and atlas segmentation was based on lung cancer patients and was validated against a national consensus dataset of 12 patients with breast cancer. The accuracy of dose predictions between manual and auto-segmented heart and substructures was evaluated by transferring the dose distribution of 240 previously treated lung cancer patients to the consensus data set. RESULTS: The hybrid auto-segmentation method performed well with a heart dice similarity coefficient (DSC) of 0.95, with no statistically significant difference between the automatic and manual delineations. The DSC for the chambers varied from 0.78-0.86 for the automatic segmentation and was comparable with the inter-observer variability. Most importantly, the automatic segmentation was as precise as the clinical experts in predicting the dose distribution to the heart and all substructures. CONCLUSION: The hybrid segmentation method performed well in delineating the heart and substructures. The prediction of dose by the automatic segmentation was aligned with the manual delineations, enabling measurement of heart and substructure dose in large cohorts. The delineation algorithm will be available for download.

An open-source nnU-net algorithm for automatic segmentation of MRI scans in the male pelvis for adaptive radiotherapy.

Background: Adaptive MRI-guided radiotherapy (MRIgRT) requires accurate and efficient segmentation of organs and targets on MRI scans. Manual segmentation is time-consuming and variable, while deformable image registration (DIR)-based contour propagation may not account for large anatomical changes. Therefore, we developed and evaluated an automatic segmentation method using the nnU-net framework. Methods: The network was trained on 38 patients (76 scans) with localized prostate cancer and tested on 30 patients (60 scans) with localized prostate, metastatic prostate, or bladder cancer treated at a 1.5 T MRI-linac at our institution. The performance of the network was compared with the current clinical workflow based on DIR. The segmentation accuracy was evaluated using the Dice similarity coefficient (DSC), mean surface distance (MSD), and Hausdorff distance (HD) metrics. Results: The trained network successfully segmented all 600 structures in the test set. High similarity was obtained for most structures, with 90% of the contours having a DSC above 0.9 and 86% having an MSD below 1 mm. The largest discrepancies were found in the sigmoid and colon structures. Stratified analysis on cancer type showed that the best performance was seen in the same type of patients that the model was trained on (localized prostate). Especially in patients with bladder cancer, the performance was lower for the bladder and the surrounding organs. A complete automatic delineation workflow took approximately 1 minute. Compared with contour transfer based on the clinically used DIR algorithm, the nnU-net performed statistically better across all organs, with the most significant gain in using the nnU-net seen for organs subject to more considerable volumetric changes due to variation in the filling of the rectum, bladder, bowel, and sigmoid. Conclusion: We successfully trained and tested a network for automatically segmenting organs and targets for MRIgRT in the male pelvis region. Good test results were seen for the trained nnU-net, with test results outperforming the current clinical practice using DIR-based contour propagation at the 1.5 T MRI-linac. The trained network is sufficiently fast and accurate for clinical use in an online setting for MRIgRT. The model is provided as open-source.

Open-source distributed learning validation for a larynx cancer survival model following radiotherapy.

INTRODUCTION: Prediction models are useful to design personalised treatment. However, safe and effective implementation relies on external validation. Retrospective data are available in many institutions, but sharing between institutions can be challenging due to patient data sensitivity and governance or legal barriers. This study validates a larynx cancer survival model performed using distributed learning without any sensitive data leaving the institution. METHODS: Open-source distributed learning software based on a stratified Cox proportional hazard model was developed and used to validate the Egelmeer et al. MAASTRO survival model across two hospitals in two countries. The validation optimised a single scaling parameter multiplied by the original predicted prognostic index. All analyses and figures were based on the distributed system, ensuring no information leakage from the individual centres. All applied software is provided as freeware to facilitate distributed learning in other institutions. RESULTS: 1745 patients received radiotherapy for larynx cancer in the two centres from Jan 2005 to Dec 2018. Limiting to a maximum of one missing value in the parameters of the survival model reduced the cohort to 1095 patients. The Harrell C-index was 0.74 (CI95%, 0.71-0.76) and 0.70 (0.66-0.75) for the two centres. However, the model needed a scaling update. In addition, it was found that survival predictions of patients undergoing hypofractionation were less precise. CONCLUSION: Open-source distributed learning software was able to validate, and suggest a minor update to the original survival model without central access to patient sensitive information. Even without the update, the original MAASTRO survival model of Egelmeer et al. performed reasonably well, providing similar results in this validation as in its original validation.

Larynx cancer survival model developed through open-source federated learning.

INTRODUCTION: Federated learning has the potential to perfrom analysis on decentralised data; however, there are some obstacles to survival analyses as there is a risk of data leakage. This study demonstrates how to perform a stratified Cox regression survival analysis specifically designed to avoid data leakage using federated learning on larynx cancer patients from centres in three different countries. METHODS: Data were obtained from 1821 larynx cancer patients treated with radiotherapy in three centres. Tumour volume was available for all 786 of the included patients. Parameter selection among eleven clinical and radiotherapy parameters were performed using best subset selection and cross-validation through the federated learning system, AusCAT. After parameter selection, ß regression coefficients were estimated using bootstrap. Calibration plots were generated at 2 and 5-years survival, and inner and outer risk groups' Kaplan-Meier curves were compared to the Cox model prediction. RESULTS: The best performing Cox model included log(GTV), performance status, age, smoking, haemoglobin and N-classification; however, the simplest model with similar statistical prediction power included log(GTV) and performance status only. The Harrell C-indices for the simplest model were for Odense, Christie and Liverpool 0.75[0.71-0.78], 0.65[0.59-0.71], and 0.69[0.59-0.77], respectively. The values are slightly higher for the full model with C-index 0.77[0.74-0.80], 0.67[0.62-0.73] and 0.71[0.61-0.80], respectively. Smoking during treatment has the same hazard as a ten-years older nonsmoking patient. CONCLUSION: Without any patient-specific data leaving the hospitals, a stratified Cox regression model based on data from centres in three countries was developed without data leakage risks. The overall survival model is primarily driven by tumour volume and performance status.