Related Experiment Video
Updated: Nov 19, 2025
07:50
Author Spotlight: Scalable Drug Screening Protocol for Efficient Discovery of M. abscessus Treatments
Published on: October 25, 2024
2.1K
Artificial Intelligence-Based Screening for Mycobacteria in Whole-Slide Images of Tissue Samples
Liron Pantanowitz1,2, Uno Wu3,4, Lindsey Seigh1
1Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA, USA.
American Journal of Clinical Pathology
|February 2, 2021
Summary
A new deep learning algorithm effectively screens digitized acid-fast-stained (AFS) slides for mycobacteria. This artificial intelligence (AI) tool is more sensitive, accurate, and faster than manual methods for identifying acid-fast bacilli (AFBs).
Area of Science:
- Digital pathology
- Artificial intelligence in medicine
- Microbiology
Background:
- Acid-fast staining (AFS) is crucial for identifying mycobacteria.
- Manual screening of AFS slides can be time-consuming and prone to error.
- Digital pathology and AI offer potential improvements in diagnostic accuracy and efficiency.
Purpose of the Study:
- To develop and validate a deep learning algorithm for screening digitized AFS slides.
- To compare the performance of AI-assisted review with traditional methods for AFB detection.
- To assess the clinical utility and ease of use of the AI system for pathologists.
Main Methods:
- Developed a deep learning algorithm using 441 whole-slide images (WSIs) of AFS tissue.
- Created a web-based gallery for pathologist review of regions of interest and corresponding WSIs.
- Conducted a comparative study on 138 AFS slides using AI-assisted analysis, manual light microscopy, and WSI evaluation without AI.
Main Results:
- The algorithm achieved an area under the curve of 0.960 at the image patch level.
- AI-assisted reviews identified significantly more AFBs compared to manual microscopy or WSI examination (P < .001).
- AI-assisted reviews demonstrated higher sensitivity, negative predictive value, accuracy, and better concordance with original diagnoses, while being less time-consuming and easier to use.
Conclusions:
- Successfully developed and clinically validated an AI-based digital pathology system for AFB screening.
- AI assistance enhances sensitivity and accuracy in detecting AFBs in anatomic pathology.
- The AI system streamlines the screening process, offering a more efficient and user-friendly alternative for pathologists.
