Machine learning and machine teaching in histopathology

Area of Science:

• Digital Pathology
• Artificial Intelligence in Medicine
• Histopathology Training

Background:

• Histopathological classification of large bowel biopsies is crucial for diagnosis.
• Medical student training in digital pathology requires effective methods.
• Artificial intelligence (AI) offers potential for enhancing diagnostic accuracy.

Purpose of the Study:

• To evaluate the effectiveness of an AI platform in training medical students to classify whole slide images (WSIs) of large bowel biopsies.
• To assess the improvement in diagnostic accuracy of medical students over multiple training rounds using AI-assisted WSI analysis.

Main Methods:

• An AI platform was trained by a consultant histopathologist to classify large bowel biopsy WSIs.
• Six medical students underwent six rounds of training, classifying five WSI cases per round.
• Student accuracy was compared against AI-generated classifications and tracked over training rounds.

Main Results:

• Student diagnostic accuracy significantly improved from a mean of 13.7% in round one to 77.1% in round six (p=0.0011).
• The AI platform facilitated learning, with students deducing morphological features that were largely accurate.
• Individual learning curves varied, with some students demonstrating faster improvement than others.

Conclusions:

• AI-powered training platforms can significantly enhance medical students' accuracy in diagnosing large bowel biopsies from WSIs.
• The study demonstrates the potential of AI as a valuable tool in histopathology education and diagnostic skill development.
• Morphological feature recognition, aided by AI, is a key component in improving diagnostic performance in digital pathology.