Researchers funded by the National Institutes of Health have developed an artificial intelligence interface that enables patients with epilepsy to communicate in real time with brain monitoring devices. A clinical study at the University of Pennsylvania Hospital involving 13 patients found the system reduced false alarms by approximately 80 percent.
Patients in the study used a smartphone application to receive alerts and notifications based on their electroencephalography (EEG) data and to submit queries to the AI system about their own brain activity. The system responded accurately to over 92 percent of questions.
The system learned from patient feedback. Patients indicated whether the device's seizure detections were accurate, false, or uncertain. Using this input, the system improved its ability to pinpoint actual seizures while reducing its false alarm rate.
Electroencephalography devices record electrical brain activity and help doctors diagnose epilepsy and detect seizure risk. The devices continuously collect detailed data, but patients generally receive information from medical experts only daily or less frequently. EEG algorithms automatically flag signals but frequently mislabel benign activity, requiring experts to manually disregard false positives.
The researchers are planning larger studies for outpatient use and preparing to deploy the system with deep brain stimulation devices for patients with Parkinson's disease. Several patients in the initial study asked if they could take the system home, which Zack Goldblum, a graduate researcher involved in the study, called "one of the best signals we could receive."