CheXtriev: Anatomy-Centered Representation for Case-Based Retrieval of Chest Radiographs
Artificial IntelligenceComputer Vision and Pattern RecognitionMachine Learning
Summary
The authors created CheXtriev, a new method to find similar chest X-ray images by focusing on specific body parts rather than looking at the whole image at once. They use a technique called graph transformers to understand how different areas of the chest relate to each other and to medical conditions. This makes the search more accurate, especially for rare findings. Their approach works better than previous methods and helps doctors find relevant images more efficiently.
Authors
Naren Akash, Arihanth Tadanki, Jayanthi Sivaswamy
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.