Hong Kong Med J 2025;31(Suppl 1):S8-10
Artificial intelligence for prostate cancer detection and classification on magnetic resonance imaging: abridged secondary publication
P Cao1, V Vardhanabhuti1, W Lam2
1 Department of Diagnostic Radiology, The University of Hong Kong, Hong Kong SAR, China
2 Department of Urology, The University of Hong Kong, Hong Kong SAR, China
 
 
  1. We used CapsuleNet for prostate lesion detection and classification via the Prostate Imaging Reporting and Data System, incorporating relative spatial information and the clinical context of lesions in relation to various anatomical structures.
  2. Deep learning methods for CapsuleNet classification have only achieved satisfactory outcomes. To improve outcomes, we used MiniSegCaps, an end-to-end network that integrates classification and segmentation, specifically designed for a small dataset.
  3. MiniSegCaps demonstrated impressive performance. We also developed a graphical user interface to illustrate its integration with the clinical workflow.