Multi-Modal Wristband Sensors May Predict Seizures

December 5, 2020
Marco Meglio
Marco Meglio

Marco Meglio, Associate Editor for NeurologyLive, has been with the team since October 2019. Follow him on Twitter @marcomeglio1 or email him at mmeglio@neurologylive.com

The initial study results may provide the basis for future evaluation as a step towards patient empowerment and objective epilepsy diagnostics for broad application.

Multi-modal wristband sensor data from easy-to-use, non-invasive devices in combination with deep learning may provide statistically significant and clinically useful seizure forecasting, according to a study presented at the American Epilepsy Society (AES) Annual Meeting, December 4–8, 2020.1

Lead author Christian Meisel, MD, PhD, department of neurology, Universitätsmedizin Berlin, and Berlin Institute of Health, and colleagues applied deep learning networks such as long short-term memory (LSTM) and 1DConv on multi-modal wristband sensor data from 69 persons with epilepsy (PWE) to assess its capability to forecast seizures. Notably, the data for the study was collected by contributions from Tobias Loddenkemper, MD, director, clinical epilepsy research, Boston Children’s Hospital.

Using evaluations based on sensitivity, time in warning, and improvement over chance (IoC), results showed that the seizure forecasting was significantly better than chance for 43.5% of patients (30 of 69 patients), yielding a mean IoC of 28.5 (±2.6) and a mean sensitivity of 75.6 (±3.8). Researchers also noted that the mean prediction horizon was 1896 (±101) seconds, a period that may be long enough to afford reasonable warning of seizures in advance.

"Our study was motivated by the potential benefits for patients and clinicians that seizure risk assessments or seizure forecasting may have. These benefits have long been known. If you ask a patient with epilepsy what they’re most concerned about within their disease, they will usually tell you that it’s the unpredictability of seizures,” Meisel told NeurologyLive.

READ MORE: Fenfluramine Demonstrates Efficacy in Lennox-Gastaut Syndrome in Phase 3 Results

Each patient within the study was equipped with the Empactica E4 wireless multi-sensor device, which recorded temperature, photoplethysmography, electrodermal activity and actigraphy, all during long-term, continuous video-electroencephalography (EEG) monitoring.

The E4 wristband is a medical grade wearable device that offers real-time physiological data acquisition, enabling researching to conduct in-depth analysis and visualization. It is equipped with multiple different sensors, including a PPG sensor which measures blood volume pulse (BVP), from which heart rate variability can be derived, and an EDA sensor, which measures the constantly fluctuating changes in certain electrical properties of the skin.2

The device also has a 3-axis accelerometer which captures motion-based activity, an infrared thermopile to read peripheral skin temperature, event mark button which tags events and links them to physiological signals, and an internal real-time clock.

Meisel and colleagues applied a leave-1-subject-out cross validation approach where matched pre-/interictal data were used for training, and testing was done on the remaining out-of-sample patient dataset.

Additional control analyses using time-matched seizure surrogate data showed that forecasting seizures was not simply based on time of day or vigilance state. To better understand how each sensor plays a role into predicting seizures, the researchers performed analyses by removing each sensor’s data individually, which indicated that all data streams contributed to seizure forecasting.

While the study included 69 PWEs, prediction performance increased with size of the training dataset, which may point to future studies that include larger datasets.

"These initial results may provide the basis for future re-evaluation, algorithm improvement and benchmarking as a step towards patient empowerment and objective epilepsy diagnostics for broad application,” Meisel et al concluded.

For more coverage of AES 2020, click here.

REFERENCES
1. Meisel C, El Atrache R, Jackson M, et al. Machine learning from wristband sensor data for wearable, non-invasive seizure forecasting. Presented at AES 2020 Annual Meeting; December 4–8, 2020; Abstract 241
2. E4 Wristband. Empatica. https://www.empatica.com/research/e4/. Accessed December 3, 2020.

Related Content:

AES | Epilepsy | Conferences | News | Clinical