An AI-assisted tool for efficient prostate cancer diagnosis in low-grade and low-volume cases

Deep-learning-based assistive tools are becoming an integral part of pathology clinics. They promise to reduce pathologists’ workloads and improve patients’ outcomes. This study focuses on a challenging task rather than an easy one to enhance these promises. It develops a deep-learning-based pipeline detecting low-grade and low-volume prostate cancer that can be easily overlooked, potentially resulting in missed therapeutic opportunities. The pipeline detects the few low-grade cancerous components within a low volume of prostate tissue and highlights high-risk regions for detailed analysis by pathologists. It can help early diagnosis of prostate cancer and effective use of other therapeutic tools at early stages, like active surveillance, rather than aggressive treatments eligible at later stages. Besides, this study conducts an external validation on data of patients from different ancestries, which can be a critical factor in the success of deep-learning-based assistive tools.