AnatomiX, an Anatomy-Aware Grounded Multimodal Large Language Model for Chest X-Ray Interpretation

Anees Ur Rehman Hashmi, Numan Saeed, Christoph Lippert · 2026

Read AnatomiX, an Anatomy-Aware Grounded Multimodal Large Language Model for Chest X-Ray Interpretation by Anees Ur Rehman Hashmi, Numan Saeed, Christoph Lippert (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-2601-03191.txt.

Description

Anees Ur Rehman Hashmi, Numan Saeed, and Christoph Lippert introduce AnatomiX, a multimodal language model that reads chest X-rays with explicit attention to anatomy. Medical models have improved at interpreting these scans, but they still stumble on spatial reasoning and often ground phrases in the wrong structures. AnatomiX follows the radiologist's workflow in two stages.

First it identifies anatomical structures and extracts their features; then a large language model uses that grounding for tasks such as phrase grounding, report generation, visual question answering, and image understanding. Across multiple benchmarks, AnatomiX improved performance on anatomy and phrase grounding, grounded diagnosis, and grounded captioning by more than twenty-five percent over existing approaches. The authors frame true anatomical correspondence, not just better scores, as the key to trustworthy medical AI.

Catalog record

CREATORAnees Ur Rehman Hashmi, Numan Saeed, Christoph Lippert
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-03191

Browse the full library · About the Tower of Babel