Dec. 28, 2021
Mayo Clinic researchers have provided the first clear evidence that a device worn on the wrist during normal daily activities can forecast seizures. The preliminary study indicates the feasibility of developing a noninvasive way to predict seizure onset for individuals with medically refractory epilepsy.
"Many people with epilepsy don't want an invasive device if they can avoid it," says Benjamin (Ben) H. Brinkmann, Ph.D., a biomedical engineer at Mayo Clinic in Rochester, Minnesota. "Our study shows that providing reliable seizure forecasts for people living with epilepsy is feasible without directly measuring brain activity."
The study — written in conjunction with international collaborators and published in the November 2021 issue of Scientific Reports — found that patterns could be identified in individuals who wore a monitoring device for 6 to 12 months while going about their daily lives. Five of 6 patients who were analyzed in the study achieved seizure alerts that were significantly more accurate than a chance predictor.
"The device gave people on average 30 minutes of warning before a seizure occurred," Dr. Brinkmann says. "We need additional studies in more patients. But we hope our research eventually helps integrate seizure forecasting into clinical practice."
Deep learning
Many people with epilepsy continue to experience seizures despite medical or surgical treatment. Seizures eventually recur in at least half of the patients who have resective surgery for epilepsy, and neuromodulation devices rarely provide long-term seizure freedom.
People living with epilepsy consistently report the unpredictability of seizures to be the most limiting aspect of their conditions. "Reliable seizure forecasts could potentially allow people living with recurrent seizures to modify their activities, take a fast-acting medication or increase neuromodulation therapy to prevent or manage impending seizures," Dr. Brinkmann says.
In the Mayo Clinic study, the wrist-worn devices captured accelerometry data as well as body temperature, blood flow, heart rate and information about the electrical characteristics of the wearer's skin. The researchers analyzed the data using long-short term memory deep learning. Participants in the study already had implanted responsive neurostimulation devices, which provided confirmation of seizures and allowed the researchers to measure the accuracy of seizure forecasting by the wrist-worn device.
"We were able to successfully forecast about two-thirds of people's seizures," Dr. Brinkmann says. "This was a preliminary feasibility study. We are gathering more data from patients and are working to increase the accuracy of the machine-learning algorithm."
Routine collaboration between clinicians and researchers allows Mayo Clinic to pursue this type of innovative research.
"As engineers, we can speak with clinicians and understand which problems we should focus on to help patients," Dr. Brinkmann says. "The partnership between the clinical and research practices is incredibly valuable."
For more information
Nasseri M, et al. Ambulatory seizure forecasting with a wrist-worn device using long-short term memory deep learning. Scientific Reports. 2021;11:21935.