Enabling clinical use of foundation models for computational pathology

TL;DR

This paper presents a method to improve the robustness and accuracy of computational pathology models by introducing novel training losses, reducing bias from technical variability without retraining foundation models.

Contribution

It introduces robustness losses during downstream training to mitigate biases and improve model generalization in clinical pathology applications.

Findings

Enhanced robustness to technical variability in pathology models.

Improved classification accuracy focusing on biologically relevant features.

Validated on over 27,000 whole-slide images from diverse patients.

Abstract

Foundation models for computational pathology are expected to facilitate the development of high-performing, generalisable deep learning systems. However, in addition to biologically relevant features, current foundation models also capture pre-analytic and scanner-specific variation that bias the predictions made by downstream task-specific models trained on these features. Here we show that introducing novel robustness losses during downstream model training reduces sensitivity to technical variability. A purpose-designed comprehensive experimentation setup with 27,042 whole-slide images from 6,155 patients is used to train thousands of models from the features of eight well-known foundation models for computational pathology. In addition to a substantial improvement in robustness, our approach improves classification accuracy by focusing on biologically relevant features. It mitigates robustness limitations of foundation models for computational pathology without retraining the foundation models themselves, enabling development of models that are more suitable in real-world clinical use.