"Conversational AI for early detection of brain health abnormalities" Introducing the core technology for next-generation healthcare services

[by Yu, Suin] A new era is expected to soon dawn where early detection of stroke-related health abnormalities and timely interventions may be achievable through simple interaction with an artificial intelligence (AI) model on a smartphone.

Professor Seo Woo-keun from the Department of Neurology at Samsung Medical Center announced on November 24 that he presented the underlying technology for a future healthcare service at the ‘IN SILICO to Patient" symposium, which took place on November 14.

The study was carried out over a three-year period by a multi-institutional consortium led by Samsung Medical Center, with participation from Korea University Ansan Hospital, Seoul Asan Medical Center, Hanyang University Guri Hospital, and Incheon National University.

The underlying technology integrates biometric health signal analysis powered by physical AI with a Large Language Model (LLM), enabling users to receive interactive consultations using only a smart device. Through this approach, individuals can conveniently access functions such as pulse monitoring and pronunciation abnormality detection without requiring any additional external equipment.

Notably, the study has demonstrated the potential for application as a medical device by incorporating sensor technologies like photoplethysmography-based biosignal monitoring and a multimodal cerebrovascular disease prediction model capable of integrating and analyzing medical data, including electronic health records (EHRs) and MRI images.

During the development of the health data model, the research team placed particular emphasis on eliminating hallucination errors, a major limitation associated with existing LLMs.

By incorporating an optimized data management model into the underlying technology, the system demonstrated no hallucination occurrences in more than 300 iterative evaluations, heightening expectations regarding the feasibility of achieving zero hallucinations.

Professor Seo Woo-keun of the Department of Neurology at Samsung Medical Center, who served as the principal investigator, remarked, "The central philosophy of this study is 'Care me at home,' illustrating how state-of-the-art AI technology can be integrated into patients' everyday lives. Our efforts were devoted to identifying practical solutions that enable the application of the developed technology to actual patients."

In fact, the research team involved patient and consumer groups as core partners from the early stages of planning. The implementation of a customized service, achieved by continuously incorporating their feedback throughout the design and validation process, is regarded as a major differentiating factor compared with existing technologies.

"This research represents the collective expertise of multiple institutions led by Samsung Medical Center. Clinical and biosignal data from Korea University Ansan Hospital and Hanyang University Guri Hospital, ethical and legal guidance from the ELSI team at Seoul Asan Medical Center, and AI modeling carried out by Incheon National University collectively enhanced the practical applicability and the predictive accuracy of the model," Seo stated.

He further noted, "The combined contributions of these institutions enabled the technology to be meaningfully integrated into patients' daily lives," underscoring the importance of the multi-institutional collaboration.