Clinically Relevant Latent Space Embedding of Cancer Histopathology Slides through Variational Autoencoder Based Image Compression

TL;DR

This paper presents a VAE-based method for highly efficient compression of cancer pathology slides that preserves clinical relevance, enabling better data interpretation and potential acceleration of clinical image search.

Contribution

The paper introduces a novel VAE training approach achieving superior compression ratios for cancer slides while maintaining clinical accuracy, and explores the interpretability of the resulting embeddings.

Findings

Achieved 1:512 compression ratio surpassing SOTA

Generated meaningful embeddings for clinical interpretation

Demonstrated potential for faster clinical image retrieval

Abstract

In this paper, we introduce a Variational Autoencoder (VAE) based training approach that can compress and decompress cancer pathology slides at a compression ratio of 1:512, which is better than the previously reported state of the art (SOTA) in the literature, while still maintaining accuracy in clinical validation tasks. The compression approach was tested on more common computer vision datasets such as CIFAR10, and we explore which image characteristics enable this compression ratio on cancer imaging data but not generic images. We generate and visualize embeddings from the compressed latent space and demonstrate how they are useful for clinical interpretation of data, and how in the future such latent embeddings can be used to accelerate search of clinical imaging data.