Related Experiment Video
Updated: Sep 17, 2025
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
269
Deep learning quantifies pathologists' visual patterns for whole slide image diagnosis
Tianhang Nan1, Song Zheng2,3, Siyuan Qiao4
1College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China.
Nature Communications
|July 2, 2025
Summary
This study introduces a new AI system that uses eye-tracking to learn from pathologists, significantly reducing annotation time for whole slide images (WSI) and improving diagnostic accuracy.
Area of Science:
- Digital pathology
- Artificial intelligence in medicine
- Computational pathology
Background:
- Pixelwise manual annotation of whole slide images (WSI) is crucial for training deep learning models in diagnostic pathology.
- Current annotation methods are time-consuming, burdening pathologists and limiting dataset size and model precision.
Purpose of the Study:
- To develop a method for acquiring pathologist expertise with minimal workload.
- To design a deep learning system that decodes visual patterns for precise WSI diagnostics.
Main Methods:
- Collected pathologist image review patterns using eye-tracking devices.
- Developed the Pathology Expertise Acquisition Network (PEAN), a deep learning system utilizing visual patterns.
- Evaluated PEAN on 5881 WSIs across 5 skin lesion categories.
Main Results:
- Eye-tracking reduced WSI annotation time to 4% of manual annotation.
- PEAN achieved an area under the curve of 0.992 and 96.3% accuracy in diagnostic prediction.
- The system effectively learns from pathologist diagnostic processes.
Conclusions:
- PEAN offers efficient data annotation, significantly reducing pathologist workload.
- The system enables precise diagnostics, aiding large-scale data collection and clinical care.
- This approach addresses limitations in current models by incorporating pathologist diagnostic processes.
