Overview
This case study outlines the deployment of a licensed Sensing AI Agent to support seizure research in patients with focal aware epilepsy. The goal is to generate a real-time, ecologically valid dataset that combines synchronized brain activity and behavioral data during seizure events. Unlike traditional clinical methods—which often fail to capture in-the-moment neurological and contextual signals—this approach enables personalized neuromarker discovery for each patient, with ongoing AI analysis to surface repeatable patterns.
Study Design
Patients wear the Muse Athena headset, a lightweight neurotechnology device that records both EEG and fNIRS signals throughout the day. When a seizure begins, the patient taps a button in a customized mobile app, developed specifically for this study. This action initiates timestamped, real-time streaming of neurobiological data to the platform, marking the start of the seizure event.
The data collection protocol follows three stages:
- Seizure Event Phase
During the seizure, EEG and fNIRS signals are continuously transmitted without patient interaction. This provides an uninterrupted, objective record of electrical and hemodynamic brain activity aligned precisely with seizure onset. - Post-Seizure Behavioral Session
Once the event has passed, the Sensing AI Agent initiates a guided conversation with the patient to collect behavioral, emotional, and environmental data—such as exposure to light, sound, movement, recent food or drink, and stress levels. These inputs are timestamped and mapped to the physiological data collected during the event. - Recovery Monitoring
In the hours following the event, the agent may prompt the patient for additional feedback to assess fatigue, confusion, emotional state, or other post-ictal effects.
AI-Driven Neuromarker Discovery
All collected data—EEG, fNIRS, behavioral inputs, and biometric signals—are processed using AI models trained to identify repeatable neuromarkers and correlate them with context and triggers. Over time, the system builds a personalized seizure profile for each participant, enabling:
Identification of unique physiological patterns preceding or accompanying seizures
Correlation of environmental and behavioral factors with seizure activity
Longitudinal analysis to support personalized forecasting or intervention strategies
This layer of AI-driven interpretation extends the value of the dataset beyond episodic reporting, offering structured insights for research and potential clinical applications.
Research Collaboration
The licensed partner is actively seeking collaboration with epilepsy research centers and neuroscience programs interested in seizure prediction, environmental trigger modeling, or real-time sensing. The goal is to share real-world datasets that include synchronized brain and behavior data during live seizure events—something historically difficult to obtain using traditional methods.
Scientific Rationale and Supporting Research
This approach is grounded in a growing body of literature highlighting the importance of multimodal, synchronized data in understanding seizure dynamics:
Lehnertz et al. (2007) demonstrated the predictive value of EEG when captured over longer timeframes, emphasizing individualized seizure signatures.
Cook et al. (2013) showed that implanted neurodevices can provide seizure forecasting, but noted scalability and invasiveness as barriers.
Nguyen et al. (2019) and Mishra et al. (2020) explored the potential of fNIRS as a non-invasive, hemodynamic marker of seizure onset.
Fisher et al. (2014) called for greater inclusion of behavioral and contextual factors in seizure research, recognizing the variability of triggers across individuals.
Davis et al. (2021) proposed the integration of wearable biosignals and AI to study patient-specific seizure pathways in ambulatory settings.
This case study responds directly to these findings by combining wearable, non-invasive sensing with AI-driven behavioral context collection in a scalable, real-world deployment.