Researchers from the University of Kansas announced on June 16 the development of a privacy-preserving artificial intelligence model designed to protect sensitive personal information in electrocardiogram (ECG) data. The new model, called PP-VAE, aims to allow sharing of clinically useful ECG information while reducing exposure of details such as age, sex, and demographic characteristics.
Fairuz Shadmani Shishir, a doctoral student in electrical engineering and computer science at the University of Kansas who led the study, said, "Modern AI systems may infer sensitive traits from ECG signals, including approximate age ranges and other personal soft-biometric information from the signals. Our goal was to develop a method that preserves clinically useful information in ECGs while reducing the exposure of sensitive personal attributes such as age, sex and demographic details."
The research team detailed their approach in Scientific Reports. Shishir said that protecting patient privacy is essential when sharing medical data between companies and institutions: "Our goal was to enable secure sharing of clinically useful ECG information without unnecessarily exposing sensitive personal attributes." The researchers used independent convolutional neural networks models to reduce identifiability while retaining predictions for conditions like left ventricular hypertrophy and five-year mortality risk.
"We proposed an AI-driven model that analyzes ECG signals to predict clinically important outcomes such as left ventricular ejection fraction (LVEF), which is an indicator of heart abnormalities and early mortality risk," Shishir said. "At the same time, the model is designed to reduce the exposure of sensitive biometric information, including age, sex and demographic characteristics derived from ECG signals." According to Shishir's co-authors Sumaiya Shomaji, Amit Noheria, Christopher Harvey, and Amulya Gupta—all affiliated with KU or KU Medical Center—the method could help hospitals share data safely for collaboration without compromising privacy.
Shishir said their experiments showed competitive performance compared with other machine-learning approaches: "We demonstrated that our model has competitive performance compared with other machine-learning approaches. The model performs well in predicting heart disease and early mortality risk while revealing less biometric information from ECG signals." He also noted efforts made by including balanced representation among male/female patients as well as racial groups: "Bias is an important issue to address... In our models we aimed to include balanced proportions... This was one way we attempted to minimize bias."
Looking ahead, researchers plan broader training on global datasets for better generalization across populations. They also intend for public release: "First, the model is designed to generalize across patients in United States. Second, we plan to make the model publicly available so anyone can use it... Institutions will be able to use our model and potentially build their own versions trained on their own datasets. Our goal is to release the model publicly in the future," Shishir said.
