Automated Patient Identity Verification Framework for Multimodal Medical Imaging Using Deep Metric Learning and Domain Adaptation

An automated multimodal patient identity verification framework using deep metric learning with DADA achieves AUCs from 0.9989 to 0.9999 on chest radiograph–CT scout evaluations.

• Why it matters: Identity inconsistencies in healthcare can threaten patient safety and data integrity, and cross-modality image differences make reliable linkage difficult.
• What they did: The method learns modality-invariant patient representations from labeled source-domain data and uses unlabeled target-domain data via data-augmented domain adaptation, excluding target labels from optimization.
• The result: In retrospectively collected routine-condition chest radiographs and CT scout images, similarity scores separate same-patient from different-patient pairs, supporting practical infrastructure for identity management and multimodal integration.

AI summary of the abstract · may contain errors · not medical advice · read the paper.