Login Search
FMDB Transactions on Sustainable Health Science Letters Submit Paper
    Crossref logo

    Archiving

    Portico

    Fernando Martins De Bulhão is a member of

    Crossref logo

Transformer-Based Multimodal Framework for Epileptic Seizure Detection, Prediction, and Clinical Decision Support

Authors:
Gnaneswari Gnanaguru, B. M. Praveen, S. Silvia Priscila

Addresses:
Department of Computer Applications, CMR Institute of Technology, Bengaluru, Karnataka, India. Department of Information Technology, Institute of Engineering and Technology, Srinivas University, Dakshina Kannada, Karnataka, India. Institute of Engineering and Technology, Srinivas University, Dakshina Kannada, Karnataka, India. Department of Computer Science, Bharath Institute of Higher Education and Research, Chennai, Tamil Nadu, India.

Abstract:

Epilepsy is a common chronic neurological condition with recurrent unprovoked seizures that affects millions of people worldwide. Pinpointing and forecasting such events accurately is essential for patient safety and clinical decision-making. Classical deep learning models, e.g., Convolutional Neural Networks (CNNs), have limitations in modelling long-term temporal dependencies in physiological data. In this paper, we present a Transformer-based Multimodal System (TME) to fuse Electroencephalogram (EEG) and Electrocardiogram (ECG) signals, effectively improving seizure detection and prediction performance. The model can learn complex global dependencies within time-series data using the self-attention mechanism. The dataset includes 469 individual data instances from standard clinical recordings. The framework is developed in Python, with PyTorch as the main deep learning library and Scikit-learn for evaluation metrics. Results show that multimodal fusion can outperform unimodal baselines. Additionally, there is a Clinical Decision Support component that enhances the physician's interpretability. Significantly better performance indicators are reported for the proposed design, indicating its potential as a reliable tool for real-time monitoring and automated diagnosis in clinical settings.

Keywords: Seizure Prediction; Decision Support; Seizure Detection; Multimodal Fusion; Automated Diagnosis; Clinical Recordings; Physiological Data; Unimodal Baselines.

Received on: 20/02/2025, Revised on: 25/04/2025, Accepted on: 08/07/2025, Published on: 08/03/2026

DOI: 10.69888/FTSHSL.2026.000591

FMDB Transactions on Sustainable Health Science Letters, 2026 Vol. 4 No. 1, Pages: 1-11

  • Views : 35
  • Downloads : 8
Download PDF