Matters Arising: Near-identical images, not foundation models, explain purported re-identification of patients from medical imaging

Read full text
Authors: Justin Engelmann, Siegfried K. Wagner, Yukun Zhou, Pearse Andrew Keane
Cited by: 0
References: 4

Abstract

report that foundation models enable re-identification of patients in medical imaging datasets, given a known medical image of the same type of targeted individual. They postulate that this is due to the ability of foundation models to identify attributes such as age, sex or ethnicity of a person from said imaging and identify this as a risk stemming from medical foundation model research.

Fields of Study & Subjects

Artificial Intelligence in Healthcare and Education Privacy-Preserving Technologies in Data Face recognition and analysis

Publication details

Journal / source
npj Digital Medicine
Volume
9
Issue
1
Pages
159–159
Publication date
13 February 2026
Language
EN