HYDEN: Hyperbolic Density Representations for Medical Images and Reports

TL;DR

HYDEN introduces a novel hyperbolic density embedding method for medical image and report representation, effectively capturing semantic uncertainty and improving interpretability and performance in zero-shot medical tasks.

Contribution

It presents a new hyperbolic density embedding approach that models semantic uncertainty in medical image-text data, enhancing interpretability and accuracy.

Findings

Outperforms baseline methods in zero-shot medical tasks

Demonstrates interpretability of hyperbolic density representations

Effective modeling of semantic uncertainty in medical data

Abstract

In light of the inherent entailment relations between images and text, hyperbolic point vector embeddings, leveraging the hierarchical modeling advantages of hyperbolic space, have been utilized for visual semantic representation learning. However, point vector embedding approaches fail to address the issue of semantic uncertainty, where an image may have multiple interpretations, and text may refer to different images, a phenomenon particularly prevalent in the medical domain. Therefor, we propose \textbf{HYDEN}, a novel hyperbolic density embedding based image-text representation learning approach tailored for specific medical domain data. This method integrates text-aware local features alongside global features from images, mapping image-text features to density features in hyperbolic space via using hyperbolic pseudo-Gaussian distributions. An encapsulation loss function is employed to model the partial order relations between image-text density distributions. Experimental results demonstrate the interpretability of our approach and its superior performance compared to the baseline methods across various zero-shot tasks and different datasets.