New way to protect patient data for training AI models
A recent study in South Korea examined the use of a cryptographic scheme widely used in secure elections to protect patient data used in training AI models.
Researchers from Asan Medical Center have applied homomorphic encryption (HE) to big data collected from multiple institutes to test a predictive AI model. HE is an encryption method that allows computations to be performed on encrypted data.
FINDINGS
In a study where findings published in JMIR Medical Informatics, EMR data from over 300,000 patients from three hospitals – Asan Medical Center, Seoul National University Hospital, and Ewha Women’s University Seoul Hospital – were collected to train an AI model to predict mortality rates within 30 days after surgery. These were coded with HE.
The research shows that HE can be applied in practice to combine large amounts of data from different sources, while ensuring privacy and security.
Findings also suggest that small hospitals can use this encryption method to develop their own AI models by securely tapping into data from larger hospitals. These models are likely to have better predictive capabilities than models developed using raw data from a single institution.
WHY IT MATTERS
More than a decade after HE was proven in a dissertation in the United States, many heavily regulated industries such as finance, IT, and healthcare have explored its application to improve data privacy. Even in elections, this encryption method is increasingly used to ensure accurate poll results.
The South Korean study also looked at HE, to try to overcome current limitations in the development of AI in healthcare. While having larger, more diverse data is key to improving the accuracy and applicability of an AI model, collecting it is a challenge given strict privacy regulations.
“In this multi-center study, we used advanced HE to protect personal information leakage and data security. Moreover, HE enables operations and predictive modeling on encrypted data, which provides an ultimate solution that can fully solve problems related to personal information leakage and data security. Moreover, HE provides the ‘strongest’ security when used properly, such as in outsourced computation, where HE protects against data leakage in computation,” the researchers noted in their findings.
THE BIGGER TREND
The Korean government has supported local medical institutions in conducting AI research by collaboration. For example, the Ministry of Health and Welfare launched the Medical Data Utilization Project, which aims to help digital health researchers connect with hospitals and access their datasets. Last year, it designated five Korean hospitals as centers for safe use of medical data.