Abstract:Accurate prediction of epileptic seizures could prove critical for improving patient safety and quality of life in drug-resistant epilepsy. Although deep learning-based approaches have shown promising seizure prediction performance using scalp electroencephalogram (EEG) signals, substantial limitations still impede their clinical adoption. Furthermore, identifying the optimal preictal period (OPP) for labeling EEG segments remains a challenge. Here, we not only develop a competitive deep learning model for seizure prediction but, more importantly, leverage it to demonstrate a methodology to comprehensively evaluate the predictive performance in the seizure prediction task. For this, we introduce a CNN-Transformer deep learning model to detect preictal spatiotemporal dynamics, alongside a novel Continuous Input-Output Performance Ratio (CIOPR) metric to determine the OPP. We trained and evaluated our model on 19 pediatric patients of the open-access CHB-MIT dataset in a subject-specific manner. Using the OPP of each patient, preictal and interictal segments were correctly identified with an average sensitivity of 99.31%, specificity of 95.34%, AUC of 99.35%, and F1- score of 97.46%, while prediction time averaged 76.8 minutes before onset. Notably, our novel CIOPR metric allowed outlining the impact of different preictal period definitions on prediction time, accuracy, output stability, and transition time between interictal and preictal states in a comprehensive and quantitative way and highlighted the importance of considering both inter- and intra-patient variability in seizure prediction.

What problem does this paper attempt to address?

The paper aims to address two key issues in epilepsy seizure prediction: 1. **Developing an efficient seizure prediction model**: The research team designed a deep learning model combining Convolutional Neural Networks (CNN) and Transformer architecture to identify preictal spatiotemporal dynamic features in the electroencephalogram (EEG) signals of epilepsy patients, enabling accurate seizure prediction. 2. **Determining the Optimal Preictal Period (OPP)**: The study proposes a new method—Continuous Input-Output Performance Ratio (CIOPR)—to comprehensively evaluate the performance of prediction models under different preictal period definitions and determine the optimal preictal period length for each patient. ### Main Contributions - **Efficient Classification Model**: A CNN-Transformer model is proposed, capable of highly sensitive identification of preictal and interictal EEG signals, with early prediction capability and stable cross-subject performance. - **Detailed Evaluation of Model Behavior**: By fitting the continuous output of the classifier to a sigmoid curve, researchers can analyze the model's performance on long-term EEG data in detail, including prediction time, output stability, and the transition time from interictal to preictal state. - **CIOPR Metric**: A new performance metric, CIOPR, is developed, which comprehensively considers factors such as prediction time, output stability, and state transition time, providing a thorough evaluation of the seizure prediction system. - **Determining Optimal Preictal Period**: The impact of different preictal period definitions on system performance is demonstrated, and the CIOPR metric is used to determine the optimal preictal period length for each patient. ### Overview of Research Methods - **Dataset and Preprocessing**: The study is based on the publicly available CHB-MIT dataset, which includes EEG recordings of 23 pediatric patients. Preprocessing steps include re-referencing, filtering, etc., to improve signal quality. - **Model Training**: Data is extracted using four different preictal period lengths (60, 45, 30, and 15 minutes) and trained and tested using the Leave-One-Seizure-Out cross-validation method. - **Model Architecture**: The proposed CNN-Transformer model includes three convolutional layers for feature extraction, followed by a transformer module to capture long-range dependencies. - **Performance Evaluation**: In addition to traditional performance metrics (such as sensitivity, specificity, etc.), the CIOPR metric is introduced to evaluate the model's performance under different preictal period lengths. In summary, this study not only proposes a high-performance seizure prediction model but also provides a systematic approach to optimize the selection of preictal period length, which is crucial for improving the practicality of clinical applications.

Preictal Period Optimization for Deep Learning-Based Epileptic Seizure Prediction

A Deep Fourier Neural Network for Seizure Prediction Using Convolutional Neural Network and Ratios of Spectral Power.

Epilepsy Seizure Prediction on EEG Using Common Spatial Pattern and Convolutional Neural Network.

Convolutional neural networks for seizure prediction using intracranial and scalp electroencephalogram

Epileptic Seizure Prediction in Scalp EEG Using an Improved HIVE-COTE Model

Pre-Ictal Seizure Prediction Using Personalized Deep Learning

Pediatric Seizure Prediction in Scalp EEG Using a Multi-Scale Neural Network With Dilated Convolutions

Epileptic seizure prediction based on multiresolution convolutional neural networks

Epileptic Seizure Prediction Through Machine Learning and Spatio-Temporal Features Based Time Series Analysis of Intracranial Electroencephalogram Data

A Generalised Seizure Prediction with Convolutional Neural Networks for Intracranial and Scalp Electroencephalogram Data Analysis

Epileptic Seizure Prediction Using Attention Augmented Convolutional Network

Patient-independent Epileptic Seizure Prediction using Deep Learning Models

Human Intracranial EEG Quantitative Analysis and Automatic Feature Learning for Epileptic Seizure Prediction

Prediction For High Risk Clinical Symptoms Of Epilepsy Based On Deep Learning Algorithm

Epileptic seizure prediction using scalp electroencephalogram signals

CLEP: Contrastive Learning for Epileptic Seizure Prediction Using a Spatio-Temporal-Spectral Network

Preictal and Interictal Recognition for Epileptic Seizure Prediction Using Pre-trained 2D-CNN Models

An End-to-End Deep Learning Approach for Epileptic Seizure Prediction

A novel epileptic seizure prediction method based on synchroextracting transform and 1-dimensional convolutional neural network

Shorter Latency of Real-time Epileptic Seizure Detection via Probabilistic Prediction

Early Prediction of Epileptic Seizures Using a Long-Term Recurrent Convolutional Network