Clinically Aware Synthetic Image Generation for Concept Coverage in Chest X-ray Models

Amy Rafferty, Rishi Ramaesh, Ajitha Rajan · 2026

Read Clinically Aware Synthetic Image Generation for Concept Coverage in Chest X-ray Models by Amy Rafferty, Rishi Ramaesh, Ajitha Rajan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15525.txt.

Description

Rafferty and co-authors present CARPA, a tool that makes synthetic chest X-rays with clinical traits edited in a set way. Real datasets miss many useful trait mixes, which caps model reliability. CARPA changes the trait list while holding the anatomy fixed, then renders faithful images. Fine-tuning on these images beat older methods across seven network backbones on a held-out MIMIC-CXR test. Precision and recall rose, doubt fell, and scores were better calibrated. Two radiologists judged the images realistic and in clinical agreement.

Catalog record

CREATORAmy Rafferty, Rishi Ramaesh, Ajitha Rajan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23505
LICENSEhttp://creativecommons.org/publicdomain/zero/1.0/
RIGHTSCC0
SOURCEarXiv

Source: https://arxiv.org/abs/2603.15525

Browse the full library · About the Tower of Babel