Image Distillation for Safe Data Sharing in Histopathology

TL;DR

This paper introduces a novel image distillation method using a latent diffusion model to create human-readable synthetic histopathology images, enabling safe data sharing and maintaining high downstream task performance.

Contribution

It presents a new approach combining latent diffusion and graph analysis to generate practical, human-readable synthetic datasets for histopathology, addressing privacy and data sharing issues.

Findings

Synthetic images enable comparable classification performance to real data.

Graph community analysis effectively selects informative synthetic images.

The method improves data sharing safety without sacrificing accuracy.

Abstract

Histopathology can help clinicians make accurate diagnoses, determine disease prognosis, and plan appropriate treatment strategies. As deep learning techniques prove successful in the medical domain, the primary challenges become limited data availability and concerns about data sharing and privacy. Federated learning has addressed this challenge by training models locally and updating parameters on a server. However, issues, such as domain shift and bias, persist and impact overall performance. Dataset distillation presents an alternative approach to overcoming these challenges. It involves creating a small synthetic dataset that encapsulates essential information, which can be shared without constraints. At present, this paradigm is not practicable as current distillation approaches only generate non human readable representations and exhibit insufficient performance for downstream learning tasks. We train a latent diffusion model and construct a new distilled synthetic dataset with a small number of human readable synthetic images. Selection of maximally informative synthetic images is done via graph community analysis of the representation space. We compare downstream classification models trained on our synthetic distillation data to models trained on real data and reach performances suitable for practical application.