Beyond Labels: A Self-Supervised Framework with Masked Autoencoders and Random Cropping for Breast Cancer Subtype Classification

TL;DR

This paper introduces a self-supervised learning framework using masked autoencoders and random cropping to improve breast cancer subtype classification from histopathological images, reducing reliance on labeled data.

Contribution

It presents a novel self-supervised embedding method with masked autoencoders and random cropping, tailored for breast cancer histopathology, enhancing feature learning without extensive labels.

Findings

Achieved competitive classification accuracy on BRACS dataset.

Demonstrated effectiveness of random cropping in data augmentation.

Showed that linear probes can effectively utilize learned embeddings.

Abstract

This work contributes to breast cancer sub-type classification using histopathological images. We utilize masked autoencoders (MAEs) to learn a self-supervised embedding tailored for computer vision tasks in this domain. This embedding captures informative representations of histopathological data, facilitating feature learning without extensive labeled datasets. During pre-training, we investigate employing a random crop technique to generate a large dataset from WSIs automatically. Additionally, we assess the performance of linear probes for multi-class classification tasks of cancer sub-types using the representations learnt by the MAE. Our approach aims to achieve strong performance on downstream tasks by leveraging the complementary strengths of ViTs and autoencoders. We evaluate our model's performance on the BRACS dataset and compare it with existing benchmarks.