AI models miss rare diseases in chest X-rays, especially in subgroups

A new study published on arXiv investigates fairness in chest X-ray (CXR) classification models, focusing on how rare conditions and specific patient subgroups are missed. The research highlights that even models with acceptable overall performance can fail to identify rare diseases, particularly in demographic subgroups. By analyzing datasets like VinDr-CXR and MIMIC-CXR/CXR-LT, the study proposes methods involving subgroup-aware weighting and tail-aware thresholding to reduce false negatives for rare conditions and specific groups. AI

IMPACT Highlights critical fairness issues in medical AI, potentially impacting deployment and requiring new auditing methods for rare conditions and subgroups.

RANK_REASON Academic paper on AI fairness in medical imaging. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• MIMIC-CXR/CXR-LT

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →