Abstracts

Rationale: Artificial intelligence (AI)-based real-time monitoring of physiological signals for patients offers various potential applications, such as seizure forecasting, detection, and analyzing anti-seizure medication effects. A major issue with past live seizure monitoring devices was high false alarm rates due to common daily activities, leading patients to remove the devices. This study introduces a platform that collects, processes, and displays real-time physiological signals from a wearable device, integrated with real-time action classification to accurately distinguish daily activities.


Methods:
The platform is implemented at Boston Children’s Hospital, USA, and Paderborn University, Germany and utilizes the E4 wristband (Empatica Inc., Boston, USA). The data processing pipeline is shown in Fig. 1. Live data from the different modalities of the E4 wristband, including galvanic skin response (GSR), heart rate (HR), blood volume pulse (BVP), and accelerometry (ACC), are encrypted and streamed to a laptop via Bluetooth using the Empatica E4 streaming server. We compute two metrics on the live data for each 3-second window: data completeness and data quality¹.



Within a pilot feasibility trial, we collected live data from 5 healthy subjects (mean age 26.8 years) performing four 10-minute actions (supine resting, walking, eating, and working on a computer) using the platform. We implemented a long short-term memory (LSTM) neural network for real-time action classification. The live data, data quality metrics, and thelikelihood of each action are displayed on a password-protected user interface (UI). This UI can be accessed on other devices within the network by entering an authenticated user ID and password, including a future option for 2-factor authentication, ensuring strict data privacy.

¹Böttcher, S., et al. "Data quality evaluation in wearable monitoring." Scientific Reports, 2022




Results:
The UI, shown in Fig. 2, displays live-streamed data and quality metrics on the left, with LSTM output on the right, indicated by a blue bar for real-time actions. During data collection, BVP data quality significantly dropped for movement-related actions, especially walking. For two subjects, a loosely tied wristband during rest showed 0% BVP quality, which was corrected using live feedback, and data collection was repeated. The average classification accuracy of the algorithm for all actions was 90.8% (resting: 85%, walking: 92.6%, eating: 92.5%, working: 93.2%).




Conclusions:
We demonstrated an AI-enabled live platform using wristband data. The platform provides real-time data quality feedback, allowing for immediate signal quality correction. It accurately classifies current actions in real time, potentially controlling false alarms and offering significant potential for algorithm personalization over time. This platform will be integrated with real-time patient monitoring for seizure detection and prediction, incorporating risk factors and trigger information.




Funding:
This study was supported by the Paderborn University research award and the Epilepsy Research Fund.