Development of deep learning-based novel auto-segmentation for the prostatic urethra on planning CT images for prostate cancer radiotherapy

• DOI: 10.1007/s12194-024-00832-8
• Publication Date: 2024-08-14T18:02:16Z
• Authors: Hisamichi Takagi; Ken Takeda; Noriyuki Kadoya; Koki Inoue; Shiki Endo; Noriyoshi Takahashi; Takaya Yamamoto; Rei Umezawa; Keiichi Jingu
• Keywords: Prostatic urethra; External beam radiotherapy

Abstract Urinary toxicities are one of the serious complications radiotherapy for prostate cancer, and dose-volume histogram prostatic urethra has been associated with such in previous reports. Previous research focused on estimating urethra, which is difficult to delineate CT images; however, these studies, limited number, mainly cases undergoing brachytherapy uses low-dose-rate sources do not involve external beam radiation therapy (EBRT). In this study, we aimed develop a deep learning-based method determining position patients eligible EBRT. We used contour data from 430 localized cancer. all cases, urethral catheter was placed when planning identify urethra. 2D 3D U-Net segmentation models. The input images included bladder prostate, while output model determined prostate’s based results both coronal sagittal directions. Evaluation metrics average distance between centerlines. centerline distances models were 2.07 ± 0.87 mm 2.05 0.92 mm, respectively. Increasing number maintaining equivalent accuracy as did study suggests potential high generalization performance feasibility using learning technology