Protecting patient privacy using AI

LAWRENCE — A collaboration between researchers on the KU Lawrence and KU Medical Center campuses is protecting patient privacy using artificial intelligence to remove medically unnecessary information from electrocardiograms (ECGs).

Moderns ECGs that are enhanced with AI can contain data about a patient’s age, sex, race and other identifying information. To address potential privacy concerns raised by the inclusion of this data, a group of computer science and medical researchers teamed up to develop a privacy-preserving AI model called PP-VAE to protect personally sensitive data.

“In our study, what we tried to do is develop an AI-based ECG diagnosis system so that we can transform the raw ECG signals to a format where we only sustain the vital information related to healthcare. And we try to conceal all the information related to a person's identity so that it does not breach the privacy of the patient,” said Sumaiya Shomaji, assistant professor in the Department of Electrical Engineering & Computer Science.

The project was led by Fairuz Shadmani Shishir, a doctoral student in electrical engineering & computer science. Shishir noted that medical data sets maintained by KU Medical Center made this cross-campus collaboration a success.

“We jointly developed an AI model, which leverages the data in the KUMC server,” Shishir said. “With that we trained our model and predicted the meaningful information for cardiovascular risk analysis.”

Other collaborators on the project include Amit Noheria, professor; Christopher Harvey, research associate; and Amulya Gupta, research assistant, all from the Department of Cardiovascular Medicine at KU Medical Center.

“This project nicely shows that AI can be used for improving diagnostics, but also you can leverage AI to safeguard the privacy of diagnostic information,” Noheria said.

Watch the video above and read the full news release about the team’s approach for more information.