CheXtriev: Anatomy-Centered Representation for Case-Based Retrieval of Chest Radiographs
TL;DR
This work presents CheXtriev, a graph-based, anatomy-aware framework for chest radiograph retrieval that leverages graph transformers to extract informative features from specific anatomical regions and captures spatial context and the interplay between anatomical location and findings.
Abstract
We present CheXtriev, a graph-based, anatomy-aware framework for chest radiograph retrieval. Unlike prior methods focussed on global features, our method leverages graph transformers to extract informative features from specific anatomical regions. Furthermore, it captures spatial context and the interplay between anatomical location and findings. This contextualization, grounded in evidence-based anatomy, results in a richer anatomy-aware representation and leads to more accurate, effective and efficient retrieval, particularly for less prevalent findings. CheXtriv outperforms state-of-the-art global and local approaches by 18% to 26% in retrieval accuracy and 11% to 23% in ranking quality. The code is available at https://github.com/cvit-mip/chextriev.
