Context-aware Self-supervised Learning for Medical Images Using Graph Neural Network

TL;DR

This paper presents a novel self-supervised learning method for medical images that leverages graph neural networks to incorporate anatomical context at regional and patient levels, improving abnormality detection.

Contribution

It introduces a dual-level self-supervised framework using graph neural networks to embed anatomical relationships, enhancing medical image analysis.

Findings

Outperforms baseline methods lacking contextual modeling.

Effectively detects lung tissue abnormalities related to COVID-19.

Handles arbitrarily sized images in full resolution.

Abstract

Although self-supervised learning enables us to bootstrap the training by exploiting unlabeled data, the generic self-supervised methods for natural images do not sufficiently incorporate the context. For medical images, a desirable method should be sensitive enough to detect deviation from normal-appearing tissue of each anatomical region; here, anatomy is the context. We introduce a novel approach with two levels of self-supervised representation learning objectives: one on the regional anatomical level and another on the patient-level. We use graph neural networks to incorporate the relationship between different anatomical regions. The structure of the graph is informed by anatomical correspondences between each patient and an anatomical atlas. In addition, the graph representation has the advantage of handling any arbitrarily sized image in full resolution. Experiments on large-scale Computer Tomography (CT) datasets of lung images show that our approach compares favorably to baseline methods that do not account for the context. We use the learned embedding for staging lung tissue abnormalities related to COVID-19.