ScholarDeck

StudyNPJ digital medicine · 2025

Re-identification of patients from imaging features extracted by foundation models.

Giacomo Nebbia, Sourav Kumar, Stephen Michael McNamara … Jayashree Kalpathy-Cramer

Read paper

OphthalmologyStudy

Foundation-model imaging features re-identified patients from 46.3% of optical coherence tomography images.

Re-identification of patients from imaging features extracted by foundation models.

Giacomo Nebbia … Jayashree Kalpathy-Cramer

NPJ digital medicine · 2025

Background

Foundation models for medical imaging are a prominent research topic, but risks associated with the imaging features they can capture have not been explored.

Purpose

We aimed to assess whether imaging features from foundation models enable patient re-identification and to relate re-identification to demographic features prediction.

Results

Optical Coherence Tomography eye scans could be matched back to the individual patient

40.3%

CFP

46.3%

OCT

25.9%

chest x-rays

More results

When training a deep learning model on the re-identification task, we reported performance of 82.3%, 93.9%, and 63.7% at image level on our internal Colour Fundus Photos, Optical Coherence Tomography, and chest x-ray data.

“

Conclusion

We showed that imaging features extracted from foundation models in ophthalmology and radiology include information that can lead to patient re-identification.

Post

Re-identification of patients from imaging features extracted by foundation models.

Foundation-model imaging features re-identified patients from 46.3% of optical coherence tomography images.

Related papers