Influence of Artificial Intelligence Assistance on Gleason Grading and Prostate Cancer Detection by Uropathologists in Daily Practice: A Prospective Multicenter Study.

内容摘要

This prospective multicenter study evaluated the real-world influence of artificial intelligence (AI) assistance on Gleason grading and prostate cancer detection. Between January and June 2025, diagnostic prostate biopsies from 130 consecutive cases with suspected prostate cancer were assessed at two Dutch hospitals. Each biopsy entry was reviewed by one of three uropathologists in three sequential steps: (1) pathologist without AI (reference standard), (2) AI (Paige Prostate Suite), and (3) pathologist with AI assistance. We assessed agreement in Gleason grading and the derived International Society of Urological Pathology grade groups (GG), tumor detection, and changes in diagnostic confidence. Tumor was found in 64.3% (182/283) of entries, with all GG observed. GG assessments differed in 16.5% (30/182) of entries when comparing AI-assisted with unassisted evaluations. Of these, 80.0% (24/30) involved a one-point change in GG, and 20.0% (6/30) involved a two-point change. Notably, 36.7% (11/30) of discrepancies reflected a shift between GG1 and GG2. Tumor detection agreement was 98.6% (279/283), with differences limited to GG1 tumors. Uropathologists' confidence in Gleason grading increased from confident to high confidence. In clinical practice, AI assistance influenced Gleason grading in 16.5% of entries. GG changes were one or two points, with clinically relevant shifts between GG1 and GG2 in 36.7%. Differences in tumor detection were limited to low-grade (GG1) cases, which typically do not require active treatment. ln addition, AI support increased uropathologists' diagnostic confidence. AI assistance may be a powerful tool to ensure more accurate patient management in prostate cancer.