A machine learning model for predicting post-stroke epilepsy risk by integrating multimodal EEG-fMRI and clinical biomarkers

 No prevention protocol written!

Precisely why was this research done? Incompetence of the mentors and senior researchers not knowing previous research? That's being polite!

Look how long we've known of the problem.

• 10% seizures post stroke (19 posts to April 2017)

• 5% epileptic seizures after stroke (10 posts to April 2021)

• epileptic seizures (6 posts to December 2015)

• post-stroke epilepsy (7 posts to December 2016)

Just maybe you want your doctor to try these solutions.

Cannabidiol May Reduce Seizures by Half in Hard-to-treat Epilepsy

Or maybe the nasal spray referred to in here:

Preventing Seizure-Caused Damage to the Brain

The answers are out there, does your doctor know about them? 

Mozart may reduce seizure frequency in people with epilepsy

 

A dietary supplement dampens the brain hyperexcitability seen in seizures or epilepsy

 The latest here:

A machine learning model for predicting post-stroke epilepsy risk by integrating multimodal EEG-fMRI and clinical biomarkers

Z

Ze Wang

H

Huanhuan Liu



Teng Ma *

• Department of Neurology II, Qingdao Traditional Chinese Medicine Hospital, Qingdao Hiser Hospital Affiliated to Qingdao University, Qingdao, Shandong, China

Abstract

Objective: 

This study aimed to develop and validate a machine learning model integrating multimodal electroencephalography-functional magnetic resonance imaging (EEG-fMRI) features with clinical biomarkers for predicting post-stroke epilepsy (PSE) risk, thus providing a quantitative tool for early identification of high-risk patients.

Methods: 

A total of 365 acute stroke patients admitted to our hospital from January 2021 to June 2024 were retrospectively enrolled and randomly divided into training (n = 256) and validation (n = 109) sets in a 7:3 ratio. Demographic data, EEG parameters, multimodal MRI indices, and serum biomarkers were collected. In the training set, univariate analysis was first performed to screen relevant factors, followed by LASSO regression for variable selection. Multivariate logistic regression was ultimately used to identify independent risk factors. Based on key predictors, random forest (RF), support vector machine (SVM), and gradient boosting (GB) models were constructed using Python. Model performance was evaluated and optimized via the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). A nomogram was developed for risk visualization, and SHapley Additive exPlanations (SHAP) values were employed for interpretability analysis to quantify the direction and magnitude of feature contributions.

Results: 

No significant differences in baseline characteristics were observed between the training and validation sets (P > 0.05), confirming data comparability. Univariate and multivariate logistic regression showed that epileptiform discharge frequency (EDF), background EEG delta wave ratio (BEDWR), stroke lesion volume (SLV), National Institutes of Health Stroke Scale (NIHSS) score, and serum neuron-specific enolase (NSE) levels were independent risk factors for PSE (all P < 0.05). Among the models, RF demonstrated superior predictive performance, with AUCs of 0.892 (training set) and 0.731 (validation set). Interpretability analysis showed that the nomogram enabled individualized risk calculation. SHAP values confirmed EDF (highest mean SHAP value), NIHSS score, and lesion volume as the top three positively contributing features (higher values correlated with increased PSE risk), aligning with regression results and validating clinical rationality.

Conclusion: 

An RF model integrating multimodal data was successfully developed to effectively predict PSE risk. EDF, NIHSS score, SLV, BEDWR, and serum NSE were identified as core predictive indicators.

Introduction

Post-stroke epilepsy (PSE) was one of the most severe complications of stroke, significantly increasing mortality risk, exacerbating neurological deficits, and adversely affecting rehabilitation and quality of life (1, 2). Currently, clinical practice lacks effective tools for early and accurate identification of high-risk patients, relying primarily on retrospective clinical feature analysis with limited predictive precision and strong subjectivity (3).

Recent advances in multimodal neuroimaging and electrophysiological techniques provide new insights into the pathological mechanisms of PSE. Studies suggest that epileptiform discharge frequency (EDF) and delta wave activity on electroencephalography (EEG), along with imaging-derived markers such as stroke lesion volume (SLV) and clinical scores (e.g., National Institutes of Health Stroke Scale, NIHSS), may be closely associated with seizure risk (4). Additionally, serum biomarkers like neuron-specific enolase (NSE) indicate the role of neuronal injury in epileptogenesis (5, 6). Recent studies have explored computed tomography (CT)-based deep learning models for PSE prediction, such as an automatic deep-learning approach for predicting post-stroke epilepsy after initial intracerebral hemorrhage based on non-contrast computed tomography imaging (7), which highlights the potential of emergency imaging modalities. However, single-modality predictors exhibit limited performance. Effectively integrating multimodal data—including electrophysiological, imaging, and clinical biomarkers—for precise individualized risk stratification remains a major clinical challenge.

Machine learning, with its capacity to handle complex, high-dimensional data, demonstrates unique advantages in extracting deep features from heterogeneous sources. Therefore, this study aims to develop a machine learning-based predictive model integrating electroencephalography-functional magnetic resonance imaging (EEG-fMRI) features and key clinical biomarkers to stratify PSE risk, providing an objective and reliable tool for early high-risk identification and personalized intervention strategies.

More at link.