Interpretable EEG seizure prediction using a multiobjective evolutionary algorithm

Abstract

Seizure prediction might be the solution to tackle the apparent unpredictability of seizures in patients with drug-resistant epilepsy, which comprise about a third of all patients with epilepsy. Designing seizure prediction models involves defining the pre-ictal period, a transition stage between inter-ictal brain activity and the seizure discharge. This period is typically a fixed interval, with some recent studies reporting the evaluation of different patient-specific pre-ictal intervals. Recently, researchers have aimed to determine the pre-ictal period, a transition stage between regular brain activity and a seizure. Authors have been using deep learning models given the ability of such models to automatically perform pre-processing, feature extraction, classification, and handling temporal and spatial dependencies. As these approaches create black-box models, clinicians may not have sufficient trust to use them in high-stake decisions. By considering these problems, we developed an evolutionary seizure prediction model that identifies the best set of features while automatically searching for the pre-ictal period and accounting for patient comfort. This methodology provides patient-specific interpretable insights, which might contribute to a better understanding of seizure generation processes and explain the algorithm’s decisions. We tested our methodology on 238 seizures and 3687 h of continuous data, recorded on scalp recordings from 93 patients with several types of focal and generalised epilepsies. We compared the results with a seizure surrogate predictor and obtained a performance above chance for 32% patients. We also compared our results with a control method based on the standard machine learning pipeline (pre-processing, feature extraction, classifier training, and post-processing), where the control marginally outperformed our approach by validating 35% of the patients. In total, 54 patients performed above chance for at least one method: our methodology or the control one. Of these 54 patients, 21 ($$\approx$$ ≈ 38%) were solely validated by our methodology, while 24 ($$\approx$$ ≈ 44%) were only validated by the control method. These findings may evidence the need for different methodologies concerning different patients.

Unsupervised EEG Preictal Interval Identification in Patients with Drug-resistant Epilepsy

“…Additionally, we compared the starting time of the preictal intervals identified using unsupervised learning with the preictal intervals found using grid-search supervised learning on EEG data from the EPILEPSIAE database (refer to Table S7 and Figure S12 in Supplementary Section 6.3). Namely, there are two studies 21,22 documenting results of preictal grid-search, which also report the identification number for each patient. This allows for a more straightforward comparison, though not ideal, as we obtained seizure-specific preictal intervals only for some seizures (within the same patient) rather than a patient-specific preictal interval for all seizures.…”

“…Only lead seizures' EEG data were analysed herein. A seizure was considered lead seizure when preceded by at least 4.5 hours of seizure-free interval 22,32 . By this criterion, 162 seizures were discarded from a total of 388 seizures, leading to the 226 seizures considered herein.…”

“…It is common to group features into (i) univariate linear, capturing, for example, the characteristics of the frequency spectrum in different frequency bands, (ii) univariate nonlinear, capturing the nonlinear behaviour of the EEG, and (iii) multivariate measures, measuring brain connectivity patterns (refer to Supplementary Section 3 for more details). The frequency bands considered in univariate linear and multivariate feature extraction comprise delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), beta (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30), and gamma (30-47 Hz) 6,12,14 .…”

“…As there is no clinical definition of the preictal interval, several studies present seizure prediction models using a fixed interval (typically a value in the range of 2 to 90 minutes) 6,12 . Since 2005, concerns regarding preictal's starting time have been addressed through the evaluation of different preictal interval durations [13][14][15][16][17][18][19][20][21][22] . There are statistical approaches that typically compare the distributions of interictal and preictal intervals and there are algorithmic approaches performing a grid-search on the preictal interval 6 .…”

Metric Learning in Freewill EEG Pre-Movement and Movement Intention Classification for Brain Machine Interfaces

“…Interpretable methods have been considered to better understand brain activities using EEG. Recently, algorithms providing clinical interpretation have been introduced in seizure prediction in epilepsy (Uyttenhove, 2020;Pinto et al, 2022) and schizophrenia detection (Vázquez et al, 2021). In addition, Three-stage algorithm was introduced to classify two spatial tasks (Yi et al, 2022).…”