Interpretable Prostate Cancer Detection using a Small Cohort of MRI Images

2026-03-18

AI Summary

This study addresses the diagnostic challenges posed by small, highly heterogeneous prostate cancer lesions in MRI when relying solely on single-modality T2-weighted imaging and a limited dataset of 162 cases. To this end, an interpretable automated detection framework is developed, leveraging transfer learning (ResNet18), data augmentation, and systematic comparisons across multiple models—including Vision Transformer, CNN, and HOG+SVM. The ResNet18-based model achieves 90.9% accuracy and 95.2% sensitivity (AUC 0.905), while HOG+SVM also demonstrates strong performance (AUC 0.917). Notably, the model’s sensitivity significantly surpasses the average performance of five radiologists (67.5%, Fleiss κ = 0.524). These findings highlight the competitive efficacy of handcrafted features under data-scarce conditions, reducing reliance on multimodal data or high-end imaging equipment while enhancing clinical interpretability and diagnostic consistency.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Interpretability, Explainability, and Transparency
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Prostate cancer is a leading cause of mortality in men, yet interpretation of T2-weighted prostate MRI remains challenging due to subtle and heterogeneous lesions. We developed an interpretable framework for automatic cancer detection using a small dataset of 162 T2-weighted images (102 cancer, 60 normal), addressing data scarcity through transfer learning and augmentation. We performed a comprehensive comparison of Vision Transformers (ViT, Swin), CNNs (ResNet18), and classical methods (Logistic Regression, SVM, HOG+SVM). Transfer-learned ResNet18 achieved the best performance (90.9% accuracy, 95.2% sensitivity, AUC 0.905) with only 11M parameters, while Vision Transformers showed lower performance despite substantially higher complexity. Notably, HOG+SVM achieved comparable accuracy (AUC 0.917), highlighting the effectiveness of handcrafted features in small datasets. Unlike state-of-the-art approaches relying on biparametric MRI (T2+DWI) and large cohorts, our method achieves competitive performance using only T2-weighted images, reducing acquisition complexity and computational cost. In a reader study of 22 cases, five radiologists achieved a mean sensitivity of 67.5% (Fleiss Kappa = 0.524), compared to 95.2% for the AI model, suggesting potential for AI-assisted screening to reduce missed cancers and improve consistency. Code and data are publicly available.

Problem

• prostate cancer detection
• small dataset
• T2-weighted MRI
• interpretable AI
• data scarcity

Innovation

• interpretable AI
• small-data learning
• prostate cancer detection
• T2-weighted MRI
• transfer learning