Many real-world image recognition problems, such as diagnostic medical imaging exams, are "long-tailed" \unicodex2013 there are a few common findings followed by many more relatively rare conditions. In chest radiography, diagnosis is both a long-tailed and multi-label problem, as patients often present with multiple findings simultaneously. While researchers have begun to study the problem of long-tailed learning in medical image recognition, few have studied the interaction of label imbalance and label co-occurrence posed by long-tailed, multi-label disease classification. To engage with the research community on this emerging topic, we conducted an open challenge, CXR-LT, on long-tailed, multi-label thorax disease classification from chest X-rays (CXRs). We publicly release a large-scale benchmark dataset of over 350,000 CXRs, each labeled with at least one of 26 clinical findings following a long-tailed distribution. We synthesize common themes of top-performing solutions, providing practical recommendations for long-tailed, multi-label medical image classification. Finally, we use these insights to propose a path forward involving vision-language foundation models for few- and zero-shot disease classification.
@article{arxiv.2310.16112,
title = {Towards long-tailed, multi-label disease classification from chest X-ray: Overview of the CXR-LT challenge},
author = {Gregory Holste and Yiliang Zhou and Song Wang and Ajay Jaiswal and Mingquan Lin and Sherry Zhuge and Yuzhe Yang and Dongkyun Kim and Trong-Hieu Nguyen-Mau and Minh-Triet Tran and Jaehyup Jeong and Wongi Park and Jongbin Ryu and Feng Hong and Arsh Verma and Yosuke Yamagishi and Changhyun Kim and Hyeryeong Seo and Myungjoo Kang and Leo Anthony Celi and Zhiyong Lu and Ronald M. Summers and George Shih and Zhangyang Wang and Yifan Peng},
journal= {arXiv preprint arXiv:2310.16112},
year = {2024}
}