ConceptVAE: Self-Supervised Fine-Grained Concept Disentanglement from 2D Echocardiographies

TL;DR

ConceptVAE is a self-supervised framework that disentangles fine-grained anatomical concepts from style in 2D echocardiography images, improving interpretability and performance in medical image analysis tasks.

Contribution

It introduces a novel pre-training method that detects and separates fine-grained concepts from style without supervision, enhancing medical image understanding.

Findings

Outperforms traditional methods in retrieval and segmentation tasks

Successfully detects anatomical structures like blood pools and septum walls

Enables generation of style-distinct synthetic data for better calibration

Abstract

While traditional self-supervised learning methods improve performance and robustness across various medical tasks, they rely on single-vector embeddings that may not capture fine-grained concepts such as anatomical structures or organs. The ability to identify such concepts and their characteristics without supervision has the potential to improve pre-training methods, and enable novel applications such as fine-grained image retrieval and concept-based outlier detection. In this paper, we introduce ConceptVAE, a novel pre-training framework that detects and disentangles fine-grained concepts from their style characteristics in a self-supervised manner. We present a suite of loss terms and model architecture primitives designed to discretise input data into a preset number of concepts along with their local style. We validate ConceptVAE both qualitatively and quantitatively, demonstrating its ability to detect fine-grained anatomical structures such as blood pools and septum walls from 2D cardiac echocardiographies. Quantitatively, ConceptVAE outperforms traditional self-supervised methods in tasks such as region-based instance retrieval, semantic segmentation, out-of-distribution detection, and object detection. Additionally, we explore the generation of in-distribution synthetic data that maintains the same concepts as the training data but with distinct styles, highlighting its potential for more calibrated data generation. Overall, our study introduces and validates a promising new pre-training technique based on concept-style disentanglement, opening multiple avenues for developing models for medical image analysis that are more interpretable and explainable than black-box approaches.