Artificial intelligence in prostate cancer diagnosis: A systematic review of advances in gleason grade and PI-RADS classification

Abstract Artificial intelligence (AI) is reshaping prostate cancer (PCa) diagnosis by integrating advanced imaging modalities such as multi-parametric MRI (mpMRI) and whole-slide imaging (WSI). Leveraging convolutional neural networks (CNNs), these models have demonstrated over 90% accuracy in tasks like tumor localization, gleason grading (GG), and risk stratification. mpMRI modalities—including T2-weighted, diffusion-weighted (DWI), and dynamic contrast-enhanced (DCE) sequences—combined with the standardized prostate imaging reporting and data system (PI-RADS) framework, support consistent clinical interpretation. In pathology, AI methods such as CNNs and U-Net architectures achieve high performance in GG and tumor segmentation, with intersection over union (IoU) scores between 0.75 and 0.85. These systems reduce subjectivity and pathologist workload. Emerging technologies like vision transformers (ViTs) and natural language processing (NLP) further advance large-scale histopathology analysis and clinical data interpretation. Explainable AI contributes to model transparency, fostering clinical trust. This review highlights critical future directions, including multimodal data fusion, real-time diagnostics, image-to-text reporting, and personalized treatment planning—underscoring AI's growing role in improving diagnostic accuracy and patient outcomes in PCa care.