Adapting Self-Supervised Learning for Computational Pathology

TL;DR

This paper investigates how to adapt self-supervised learning, specifically DINOv2, for computational pathology by proposing domain-specific modifications and evaluating their effectiveness on various benchmarks.

Contribution

The study introduces tailored augmentations, regularization, and position encodings for SSL in pathology, demonstrating improved performance on domain-specific benchmarks.

Findings

Modified augmentations improve feature learning in pathology images.

Tailored regularization enhances model robustness.

Domain-specific position encodings boost downstream task accuracy.

Abstract

Self-supervised learning (SSL) has emerged as a key technique for training networks that can generalize well to diverse tasks without task-specific supervision. This property makes SSL desirable for computational pathology, the study of digitized images of tissues, as there are many target applications and often limited labeled training samples. However, SSL algorithms and models have been primarily developed in the field of natural images and whether their performance can be improved by adaptation to particular domains remains an open question. In this work, we present an investigation of modifications to SSL for pathology data, specifically focusing on the DINOv2 algorithm. We propose alternative augmentations, regularization functions, and position encodings motivated by the characteristics of pathology images. We evaluate the impact of these changes on several benchmarks to demonstrate the value of tailored approaches.