| Abstract: |
Brachytherapy is an important treatment modality for prostate cancer. Three-dimensional transrectal ultrasound (3D TRUS) is a common intraoperative imaging tool for prostate brachytherapy and deep learning (DL) techniques have been widely used to automate the analysis of 3D TRUS volumes. We present a review on the applications of convolutional neural networks (CNNs), one of the most successful DL architectures in medical imaging, in 3D US-based prostate brachytherapy. The main tasks involve prostate and organs-at-risk segmentation, multi-needle detection, MR to TRUS image registration, and dominant intraprostatic lesion detection. A detailed review of publications addressing each task is presented, summarizing the variant CNN architectures and state-of-the-art applications. After reviewing the development of CNNs in US-based prostate brachytherapy, we highlight the current achievements, address the major challenges, and discuss the future outlooks. © 2024 selection and editorial matter, Aaron Fenster; individual chapters, the contributors. |
