Deep Learning Enhances Detection of Extracapsular Extension in Prostate Cancer from mpMRI of 1001 Patients

Abstract

Extracapsular extension (ECE) is detected in approximately one-third of newly diagnosed prostate cancer (PCa) cases at stage T3a or higher and is associated with increased rates of positive surgical margins and early biochemical recurrence following radical prostatectomy (RP). This study presents the development of AutoRadAI, an end-to-end, user-friendly artificial intelligence (AI) pipeline designed for the identification of ECE in PCa through the analysis of multiparametric MRI (mpMRI) fused with prostate histopathology. The dataset consists of 1001 patients, including 510 pathology-confirmed positive ECE cases and 491 negative ECE cases. AutoRadAI integrates comprehensive preprocessing followed by a sequence of two novel deep learning (DL) algorithms within a multi-convolutional neural network (multi-CNN) strategy. The pipeline exhibited strong performance during its evaluation. In the blind testing phase, AutoRadAI achieved an area under the curve (AUC) of 0.92 for assessing image quality and 0.88 for detecting the presence of ECE in individual patients. Additionally, AutoRadAI is implemented as a user-friendly web application, making it ideally suited for clinical applications. Its data-driven accuracy offers significant promise as a diagnostic and treatment planning tool. Detailed instructions and the full pipeline are available at https://autoradai.anvil.app and on our GitHub page at https://github.com/PKhosravi-CityTech/AutoRadAI.