Related Experiment Videos
AI-assisted TURP slide review: a multi-reader study of efficiency, diagnostic performance, and human-AI error
C Brigida Pascarella1, Mariantonia Giobbe1, Elisabetta Maffei2
1Department of Medicine, Surgery, and Dentistry "Scuola Medica Salernitana", University of Salerno, Via Salvador Allende, 43, Baronissi, SA, 84081, Italy.
Virchows Archiv : an International Journal of Pathology
|July 21, 2026
Summary
Artificial intelligence (AI) significantly speeds up prostate cancer diagnosis in transurethral resection of the prostate (TURP) specimens. This AI tool enhances diagnostic accuracy and efficiency, especially for less experienced pathologists in high-volume settings.
Area of Science:
- Digital Pathology
- Artificial Intelligence in Medicine
- Prostate Cancer Diagnostics
Background:
- Transurethral resection of the prostate (TURP) pathology is a high-volume, low-prevalence diagnostic challenge.
- Accurate identification of rare malignant foci in TURP specimens is critical.
- Existing artificial intelligence (AI) tools show promise but require real-world workflow validation.
Purpose of the Study:
- To evaluate the impact of an AI tool on the diagnostic workflow for TURP specimens.
- To assess changes in pathologist review time and diagnostic accuracy with AI assistance.
- To determine the utility of AI as a decision-support tool in TURP pathology.
Main Methods:
- Development of an AI tool for slide-level detection of prostatic adenocarcinoma using H&E whole-slide images.
- Feature extraction via UNI and classification via a CLAM framework.
- Evaluation of 102 TURP slides by five pathologists with varying experience levels, comparing unaided vs. AI-assisted review.
Main Results:
- AI assistance significantly reduced per-slide review time for all readers (mean reduction of 34.5 seconds).
- Time savings were most pronounced among less experienced pathologists.
- Overall diagnostic accuracy improved from 95.3% (unaided) to 98.2% (AI-assisted), with improved accuracy in less experienced readers.
Conclusions:
- AI-assisted review enhances efficiency in TURP pathology workflows.
- AI maintains diagnostic accuracy and can improve it, particularly for less experienced pathologists.
- AI serves as a valuable workflow-oriented decision-support tool in high-volume, low-prevalence diagnostic settings like TURP pathology.