Automated Detection of Faciobrachial Dystonic Seizures Related Events in LGI1 Autoimmune Encephalitis Patients with Wearables

ObjectiveTo evaluate the potential of wrist-worn wearable devices to detect and quantify Faciobrachial Dystonic Seizures (FBDS) and related events associated with leucine-rich glioma Inactivated-1 (LGI1)-IgG autoimmune encephalitis (LGI1 AIE). MethodSeven patients and four control subjects were monitored with Empatica E4 wristbands in both hospital and ambulatory environments. The analysis focused on the pre- and post-immunotherapy signals of accelerometry (ACC), electrodermal activity (EDA), heart rate (HR), and blood volume pulse (BVP). A two-stage semi-supervised machine learning approach was developed, utilizing a proprietary algorithm and Support Vector Machine (SVM) classifier to identify FBDS-related events. ResultsSignificant differences were observed in the characteristics of signals recorded in patients with FBDS compared to controls during sleep periods. LGI1 AIE-associated abnormal events were more frequent, persisted longer, and generated higher ACC amplitude compared to post-immunotherapy and arousal events in the control group. Elevated tonic and phasic EDA were noted in patients, particularly before and after immunotherapy, with a notable decrease in mean and median EDA activity post-treatment, correlating with reduced limbic activation. No significant changes were observed in HR and BVP. SignificanceThe findings affirm the potential for accurate and automated detection of FBDS and its related events using wearable devices, offering a non-invasive method to quantify seizure burden and treatment efficacy. This approach could minimize the logistical challenges of in-hospital monitoring and provide continuous, decentralized means, improving patient care and clinical decision-making. Future research should focus on expanding the method to daytime monitoring and comparing its effectiveness with in-hospital video-EEG and EMG polygraphy. Key pointsO_LIAssessed the potential of using wearable technology to detect and monitor high-frequency Faciobrachial Dystonic Seizures (FBDS) and related events in anti-LGI1 autoimmune encephalitis. C_LIO_LIDeveloped a two-stage, automated machine learning algorithm to automatically isolate and classify FBDS-associated events using signals recorded with Empatica E4 wristbands. C_LIO_LIObserved significant differences in the wearable signals of patients with FBDS between pre- and post-immunotherapy, and between patient signals and normal arousal signals of the control group during sleep periods. C_LIO_LIDemonstrated the feasibility of wearable devices to provide objective measures of FBDS-related events, aiding in the quantification of treatment response and influencing clinical decision-making. C_LI