EECS Researchers Unveil Method for Safeguarding Privacy of Electrocardiograms

A common misperception exists that electrocardiograms (ECGs) simply contain data about heart activity. However, modern ECGs enhanced with artificial intelligence (AI) can reveal a patient’s sex, age, race, and even exact identity, raising fresh privacy concerns. To address this, researchers from the University of Kansas have developed a privacy-preserving AI model called PP-VAE that protects personally sensitive data while retaining clinically useful information.

Led by doctoral student Fairuz Shadmani Shishir and supervised by faculty member Sumaiya Shomaji, assistant professor in the Department of Electrical Engineering & Computer Science, the team created a model that analyzes ECG signals to predict important outcomes like left ventricular ejection fraction (LVEF) and early mortality risk, while simultaneously reducing the exposure of sensitive biometric traits such as age, sex, and demographic details. This innovation, detailed in Scientific Reports, aims to enable secure data sharing across healthcare organizations and research institutions without compromising patient privacy. The researchers also emphasize that their model helps reduce bias by including balanced representation across sex and racial groups, and they plan to make the model publicly available to foster wider adoption and further development.

The EECS Department at KU is at the forefront of interdisciplinary research that bridges cutting-edge technology with critical societal needs. By fostering collaborations between electrical engineering, computer science, and medical experts, the department creates an environment where innovations like this privacy-preserving AI can flourish. With a commitment to ethical AI development, data security, and real-world impact, EECS provides the foundational expertise and collaborative spirit necessary to tackle complex challenges at the intersection of healthcare, privacy, and machine learning.

For the full story, including more details on the research methodology and future applications, please visit here.