Self-training of Machine Learning Models for Liver Histopathology: Generalization under Clinical Shifts

TL;DR

This paper explores self-training methods for liver histopathology image analysis, demonstrating that student models can outperform teachers and approach fully supervised performance with fewer annotations, especially under clinical data shifts.

Contribution

It introduces a teacher-student self-training framework tailored for NASH histopathology, showing improved generalization under clinical shifts with limited annotations.

Findings

Student models outperform teachers by 3% in macro F1 score.

Self-trained models approach fully supervised performance with half the annotations.

Models maintain robustness under clinical data shifts.

Abstract

Histopathology images are gigapixel-sized and include features and information at different resolutions. Collecting annotations in histopathology requires highly specialized pathologists, making it expensive and time-consuming. Self-training can alleviate annotation constraints by learning from both labeled and unlabeled data, reducing the amount of annotations required from pathologists. We study the design of teacher-student self-training systems for Non-alcoholic Steatohepatitis (NASH) using clinical histopathology datasets with limited annotations. We evaluate the models on in-distribution and out-of-distribution test data under clinical data shifts. We demonstrate that through self-training, the best student model statistically outperforms the teacher with a $3\%$ absolute difference on the macro F1 score. The best student model also approaches the performance of a fully supervised model trained with twice as many annotations.