Isaac Shiri

Deep learning-based auto-segmentation of organs at risk in high-dose rate brachytherapy of cervical cancer

Delineation of organs at risk (OARs), such as the bladder, rectum and sigmoid, plays an important role in the delivery of optimal absorbed dose to the target owing to the steep gradient in high-dose rate brachytherapy (HDR-BT). In this work, we propose a deep convolutional neural network-based approach for fast and reproducible auto-contouring of OARs in HDR-BT. Images of 113 patients with locally-advanced cervical cancer were utilized in this study. We used ResU-Net deep convolutional neural network architecture, which uses long and short skip connections to improve the feature extraction procedure and the accuracy of segmentation. Seventy-three patients chosen randomly were used for training, 10 patients for validation, and 30 patients for testing. Well established quantitative metrics, such as Dice similarity coefficient (DSC), Hausdorff distance (HD), and average symmetric surface distance (ASSD), were used for evaluation. The DSC values for the test dataset were 95.7 ± 3.7%, 96.6 ± 1.5% and 92.2 ± 3.3% for the bladder, rectum, and sigmoid, respectively. The HD values (mm) were 4.05 ± 5.17, 1.96 ± 2.19 and 3.15 ± 2.03 for the bladder, rectum, and sigmoid, respectively. The ASSDs were 1.04 ± 0.97, 0.45 ± 0.09 and 0.79 ± 0.25 for the bladder, rectum, and sigmoid, respectively. The proposed deep convolutional neural network model achieved a good agreement between the predicted and manually defined contours of OARs, thus improving the reproducibility of contouring in brachytherapy workflow.

Deep Learning-based Non-rigid Image Registration for High-dose Rate Brachytherapy in Inter-fraction Cervical Cancer

Abstract In this study, an inter-fraction organ deformation simulation framework for the locally advanced cervical cancer (LACC), which considers the anatomical flexibility, rigidity, and motion within an image deformation, was proposed. Data included 57 CT scans (7202 2D slices) of patients with LACC randomly divided into the train (n = 42) and test (n = 15) datasets. In addition to CT images and the corresponding RT structure (bladder, cervix, and rectum), the bone was segmented, and the coaches were eliminated. The correlated stochastic field was simulated using the same size as the target image (used for deformation) to produce the general random deformation. The deformation field was optimized to have a maximum amplitude in the rectum region, a moderate amplitude in the bladder region, and an amplitude as minimum as possible within bony structures. The DIRNet is a convolutional neural network that consists of convolutional regressors, spatial transformation, as well as resampling blocks. It was implemented by different parameters. Mean Dice indices of 0.89 ± 0.02, 0.96 ± 0.01, and 0.93 ± 0.02 were obtained for the cervix, bladder, and rectum (defined as at-risk organs), respectively. Furthermore, a mean average symmetric surface distance of 1.61 ± 0.46 mm for the cervix, 1.17 ± 0.15 mm for the bladder, and 1.06 ± 0.42 mm for the rectum were achieved. In addition, a mean Jaccard of 0.86 ± 0.04 for the cervix, 0.93 ± 0.01 for the bladder, and 0.88 ± 0.04 for the rectum were observed on the test dataset (15 subjects). Deep learning-based non-rigid image registration is, therefore, proposed for the high-dose-rate brachytherapy in inter-fraction cervical cancer since it outperformed conventional algorithms.