Wearable sensors on the wrists or ankles use algorithms to provide an individualized, person-oriented approach to seizure detection, according to findings published in Epilepsia. Tobias Loddenkemper, MD, and colleagues used a convolutional neural network (CNN) long short-term memory (LSTM) model of seizure detection in 166 patients (median age, 10.0) and 13,254 hours of sensor data. Using the CNN-LSTM model, the researchers found that accelerometry (ACC), blood pulse volume (BPV), and fusion performance was better than chance; ACC and BVP data fusion attained a best detection performance of 83.9% sensitivity with a 35.3% false positive rate. The team identified 19 of 28 seizure types by at least one data modality with an area under the curve receiver operating characteristic performance of greater than 0.8. The results “contribute to a paradigm shift in epilepsy care that entails non-invasive seizure detection, provides time-sensitive and accurate data on additional clinical seizure types, and proposes a novel combination of an out-of-the-box monitoring algorithm with an individualized personoriented seizure detection approach,” Dr. Loddenkemper and colleagues wrote.
