Related Experiment Video
Updated: Oct 18, 2025
08:05
Detection and Isolation of Cancer in Prostate Biopsies Using Stimulated Raman Histology and Artificial Intelligence
Published on: June 10, 2025
776
A Hybrid Human-Machine Learning Approach for Screening Prostate Biopsies Can Improve Clinical Efficiency Without
David Dov1, Serge Assaad1, Ameer Syedibrahim1
1From the Department of Electrical and Computer Engineering (Dov, Assaad, Syedibrahim, Bell, Carin), Duke University, Durham, North Carolina.
Archives of Pathology & Laboratory Medicine
|September 30, 2021
Summary
A new hybrid human-machine learning system efficiently screens prostate biopsies. This tool prioritizes potentially malignant regions for pathologists, saving time on benign cases.
Area of Science:
- Pathology
- Artificial Intelligence
- Urology
Background:
- Accurate prostate cancer diagnosis relies on histologic review of multiple biopsies.
- Increasing pathology workload necessitates improved efficiency tools.
- Current deep learning approaches often work independently of pathologists.
Purpose of the Study:
- To develop an efficient hybrid human-machine learning system for prostate biopsy screening.
- To integrate pathologist expertise with deep learning for improved diagnostic workflow.
Main Methods:
- Developed an algorithm to identify 20 high-probability malignant regions of interest per biopsy.
- Pathologist review was limited to approximately 2% of the tissue area.
- Evaluated the system with 100 prostate biopsies reviewed by 4 pathologists.
Main Results:
- The hybrid system achieved high sensitivity (99.2%) in identifying malignant biopsies needing further review.
- Most benign biopsies (72.1%) were correctly identified as not requiring further review.
- Pathologist review time was significantly reduced by focusing on critical regions.
Conclusions:
- This novel hybrid system can efficiently triage benign prostate biopsies.
- The approach conserves pathologist time for detailed evaluation of malignant cases.
- This technology has the potential to enhance efficiency in prostate cancer diagnostics.
