Research
AI interface lets epilepsy patients communicate with monitoring devices in real time
In a small hospital study, patient feedback helped an AI system reduce its seizure monitor’s false alarm rate by 80%.
Researchers at the University of Pennsylvania tested a smartphone AI interface that let epilepsy patients ask questions about real-time EEG data and provide feedback on seizure alerts during hospital monitoring stays.
The study included 13 patients, who exchanged more than 1,300 messages with the system. Researchers reported that it answered more than 92% of questions about brain activity with accurate, relevant information. Patient feedback helped fine-tune the seizure-detecting model, reducing its false alarm rate by 80% over the study. Learning from expert feedback decreased alarms by another 2%.
The researchers said the study’s small size and hospital setting warrant further research. They are planning larger outpatient studies and are preparing to test the platform for communication between deep brain stimulation devices and patients with Parkinson’s disease.
