# Graph-of-Differences: Anatomy-Structured Difference Alignment for Medical Image Re-Identification

Source: [Graph-of-Differences: Anatomy-Structured Difference Alignment for Medical Image Re-Identification](<https://www.sota2.com/research/paper/6a3a1341206542251d7c71b5>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-24T05:00:04.551Z

Authors: Nichula Wasalathilaka, Abhijit Das, Imran Razzak, Dwarikanath Mahapatra

Published: 2026-06-19T12:18:12.000Z

## Abstract

Medical image re-identification (MedReID) enables longitudinal patient linkage but remains vulnerable to shortcut learning and often produces decisions that clinicians cannot audit against named anatomy. We propose Graph-of-Differences (GoD), which grounds identity comparisons in explicit anatomical structure. Each image is represented as an anatomy graph whose nodes correspond to named anatomical regions; given an image pair, soft node correspondence is established, and differences are computed over matched anatomy. A graph-level difference alignment objective ties these anatomy-matched differences to the global backbone difference, ensuring the retrieval signal is anchored in homologous structures rather than arbitrary spatial tokens. Explanations are defined over named graph nodes and quantitatively audited via node insertion/deletion tests, replacing unstable pixel heatmaps with verifiable structure-level evidence. On internal benchmarks, GoD improves Rank-1 by +7.1 pp on fundus and +3.1 pp on CXR over a strong frozen-backbone baseline, with further gains on zero-shot external transfers confirming that anatomy grounding improves both accuracy and generalization. Code is available at https://github.com/GenMI-Lab/GoD.git.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Medical Re-identification on CXR NIH14 (internal)](<https://www.sota2.com/research/sota/medical-re-identification-on-cxr-nih14-internal.md>): Medical Re-identification; CXR NIH14 (internal); R1: 90.33.

- [Medical Re-identification on CXR CheXpert (external)](<https://www.sota2.com/research/sota/medical-re-identification-on-cxr-chexpert-external.md>): Medical Re-identification; CXR CheXpert (external); R1: 33.59.

- [Medical Image Re-identification on ODIR Fundus](<https://www.sota2.com/research/sota/medical-image-re-identification-on-odir-fundus.md>): Medical Image Re-identification; ODIR Fundus; CMC-R1: 71.11.

- [Medical Image Re-identification on Mess2 Fundus](<https://www.sota2.com/research/sota/medical-image-re-identification-on-mess2-fundus.md>): Medical Image Re-identification; Mess2 Fundus; CMC-R1: 65.79.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.21368>)

[PDF](<https://arxiv.org/pdf/2606.21368.pdf>)