PROJECT SUMMARY Uncontrolled epilepsy is a major public health burden, affecting over 1.5 million patients in the US, resulting in total direct costs of up to $50,000 per year for individual patients. Generalized tonic-clonic seizures (GTCS) have been the focus of wearable-based seizure monitoring research for the past decade, while little work has been done in the detection of other seizure types. Focal seizures (FS) in particular, are the most common seizure type, affecting over 60% of people with active epilepsy. Accurate FS detection can promote safety, providing real-time alerting to caregivers, and can support more accurate seizure tracking, bypassing the need for manual diaries and providing important data for physicians to better manage epilepsy. Our team has developed a software application on a popular consumer wearable device, to record data from accelerometer (ACC) and photoplethysmography (PPG) biosensors, with the goal of providing an easy, non-stigmatizing method for FS monitoring. We propose a new methodology of detection that leverages the unique seizure characteristics of focal seizures to substantially reduce the rate of false alarms, which can result in poor compliance with monitoring. Our approach is based on the scientific premise that while focal seizures can vary significantly across individuals, they are usually far less variable within individuals, owing to their propensity for onset and propagation in the same symptomatogenic zones. We propose an adaptive methodology that can accurately classify FSs for specific individuals over time. We are uniquely positioned to complete this goal, as we can leverage the thousands of hours of data we have collected from previous trials, and our team has extensive prior experience in training and testing seizure detection algorithms. In this proposal, we plan to develop our algorithm through the following aims: (1) Developing a patient-independent focal seizure classification methodology leveraging data we have obtained from previous IRB approved research. (2) Enhancing our algorithm by creating a patient- dependent classification methodology. This aim, in particular, will allow the proposed algorithm to significantly reduce false alarm rates (FARs). (3) Prospectively validating the proposed algorithm in an observational study. We expect the final detector to significantly improve FAR without sacrificing sensitivity. If successful, we will submit a Phase II proposal focused on further validation, expansion to include ambulatory patients, and commercialization. Our overall goal is to use our clinical and technical expertise to significantly improve the lives of people with epilepsy. We believe that with this algorithm technology on the EpiWatch digital health platform, we can help ease the physical, mental, and financial burdens of uncontrolled epilepsy, thus improving quality-of-life for people with epilepsy.