Ambulatory seizure forecasting with a wrist-worn device using long-short term memory deep learning

 Does your doctor already have a protocol to prevent post stroke epilepsy and seizures? 

Your risk of post-stroke seizure is highest in the first 30 days following a stroke. Approximately 5 percent of people will have a seizure within a few weeks after having a stroke, according to the National Stroke Association.

Ambulatory seizure forecasting with a wrist-worn device using long-short term memory deep learning

 

• Mona Nasseri,
• Tal Pal Attia,
• Boney Joseph,
• Nicholas M. Gregg,
• Ewan S. Nurse,
• Pedro F. Viana,
• Gregory Worrell,
• Matthias Dümpelmann,
• Mark P. Richardson,
• Dean R. Freestone &
• Benjamin H. Brinkmann 

Scientific Reports volume 11, Article number: 21935 (2021) Cite this article

• 1848 Accesses

• 178 Altmetric

• Metrics details

Abstract

The ability to forecast seizures minutes to hours in advance of an event has been verified using invasive EEG devices, but has not been previously demonstrated using noninvasive wearable devices over long durations in an ambulatory setting. In this study we developed a seizure forecasting system with a long short-term memory (LSTM) recurrent neural network (RNN) algorithm, using a noninvasive wrist-worn research-grade physiological sensor device, and tested the system in patients with epilepsy in the field, with concurrent invasive EEG confirmation of seizures via an implanted recording device. The system achieved forecasting performance significantly better than a random predictor for 5 of 6 patients studied, with mean AUC-ROC of 0.80 (range 0.72–0.92). These results provide the first clear evidence that direct seizure forecasts are possible using wearable devices in the ambulatory setting for many patients with epilepsy.

Introduction

Despite optimized medication therapy, resective surgery, and neuromodulation therapy, many people with epilepsy continue to experience seizures. Half or more of patients who undergo resective surgery for epilepsy have eventual recurrence of seizures^{1, 2}, and devices for neuromodulation rarely achieve long-term seizure freedom^{3, 4}. People living with epilepsy consistently report the unpredictability of seizures to be the most limiting aspect of their condition⁵. Reliable seizure forecasts could potentially allow people living with recurrent seizures to modify their activities, take a fast-acting medication, or increase neuromodulation therapy to prevent or manage impending seizures. Accurate seizure forecasts have been demonstrated using invasively sampled ultralong-term EEG in ambulatory canine^6,7,8 and human subjects^{9,10,11,12,13,14}, including a prospective study with a dedicated device¹¹. However, invasive devices may not be acceptable for some patients with epilepsy, and no clinically available invasive device currently has the capability to sample and telemeter data needed for seizure forecasting. Hence there is presently great interest in forecasting seizures using wearable or minimally invasive devices. Deep learning approaches have shown promising performance for a variety of difficult applications¹⁵, including seizure forecasting⁷. In particular these “end-to-end learning” methods are attractive for seizure forecasting given the challenges of identifying salient features in ultra-long term time-series data, and the heterogeneity in time series data characteristics between different patients. The power and capability of deep learning algorithms trained on very large datasets hold promise to enable applications not previously believed possible, and may open the door to seizure forecasting with noninvasive sampling devices.

Many challenges exist in designing a reliable system for forecasting seizures from noninvasively recorded data. Training, testing, and validating a forecasting algorithm requires ultra-long duration recordings with an adequate number of seizures. Additionally, concurrent video and/or EEG validation of seizures in an ambulatory setting over months to years is logistically difficult, and is not possible using conventional in-hospital monitoring methods. Self-reported seizure diaries are the most accessible validation, but the poor reliability of such diaries is widely recognized^{11, 16}. Performing device studies on in-hospital patients with concurrent video-EEG validation is logistically feasible, but such studies are expensive, and limited in duration, and restrict normal daily activities which could produce false alarms, such as exercise, brushing teeth, or other activities. Because of these challenges an ILAE-IFCN working group recently published guidelines¹⁷ for seizure detection studies with non-invasive wearable devices, but few studies achieve phase 3–4 evidence in an ambulatory setting¹⁸. In studies of seizure forecasting it is imperative that ambulatory data including the full range of normal activities be included in the training, testing, and validation sets.

Seizure prediction with wearable devices was recently investigated in a cohort of in-hospital patients¹⁹ using a cross-patient deep learning algorithm on data recorded from Empatica E4 devices. The dataset was comprised of multiday recordings from 69 epilepsy patients (28 female, duration 2311.4 h, 452 seizures). In a leave-one-patient-out cross-validation approach, they achieved better than chance prediction in 43% of patients, with no difference in performance between generalized and focal seizure types. It has also been shown that seizure occurrence can be modeled as circadian or multiday patterns of seizure risk over long periods^{20, 21}, and these patterns may be used to forecast seizures²². Using a mobile electronic seizure diary application²¹ seizure forecasts calculated based on circadian and multiday seizure cycles using data from 50 application users produced accurate forecasts for approximately half the cohort. Long-term cycles of seizure risk offer complementary information to direct forecasting of seizures, and signals from wearable fitness trackers have been shown to have value in identifying circadian and multidian cycles of seizure risk²³.

This study aimed to develop a wearable seizure forecasting system for ambulatory use, and to evaluate the forecasting performance relative to seizures identified with concurrent chronic intracranial EEG (iEEG).