Potter‐Optimized Multiscale Bayesian Gated Axial Quantized Attention Network for Robust Multichannel EEG–ECG‐Based Seizure Prediction

ABSTRACT Accurate seizure prediction from physiological signals remains challenging because electroencephalography (EEG)–ECG recordings contain substantial noise, nonlinear temporal patterns, inter‐subject variability, and complex cross‐modal dependencies. To address these challenges, this study proposes a Potter‐Optimized Multiscale Bayesian Gated Axial Quantized Attention Network (PM‐BGAQAN) for robust three‐state seizure prediction using multichannel EEG–ECG signals. The framework first employs adaptive weighted mean filter (AWMF) to suppress physiological noise while preserving relevant signal characteristics, followed by Convolutional Swin‐Transformer Network (CSTN) for hierarchical multiscale spatiotemporal feature extraction and multimodal EEG–ECG representation learning. Inclusion of Bayesian Asymmetric Quantized Neural Network (BAQNN) is followed next for uncertainty‐aware quantized features, whereas multiscale gated axial reverse attention (MGARA) exploits the discriminatory power of multiscale temporal dependencies using gated axial attention. Finally, Potter Optimization Algorithm (POA) optimizes the network parameters to improve classification performance and convergence. The proposed framework is evaluated using the Siena Scalp EEG dataset with timestamp metadata and the SeizeIT2 EEG–ECG wearable dataset for classification of preictal, ictal, and interictal states. Experimental results demonstrate an overall classification accuracy of 99.86%, whereas subject‐independent validation achieves 98.06% average accuracy across the evaluated datasets. Ablation analysis further confirms the contribution of the individual components to the integrated framework. The results indicate that PM‐BGAQAN provides an effective multimodal representation‐learning strategy for robust seizure‐state prediction and establishes a basis for future real‐time, wearable, and clinically validated seizure monitoring systems.

Authors

Institutions

Publication Details

Journal
Developmental Neurobiology
Published
2026-11-14
DOI
https://doi.org/10.1002/dneu.70073
Primary Topic
EEG and Brain-Computer Interfaces
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Potter‐Optimized Multiscale Bayesian Gated Axial Quantized Attention Network for Robust Multichannel EEG–ECG‐Based Seizure Prediction

G. Shobana, K. R. Baskaran
Developmental Neurobiology
EEG and Brain-Computer Interfaces
article

Potter‐Optimized Multiscale Bayesian Gated Axial Quantized Attention Network for Robust Multichannel EEG–ECG‐Based Seizure Prediction

G. Shobana, K. R. Baskaran
article en

Abstract

ABSTRACT Accurate seizure prediction from physiological signals remains challenging because electroencephalography (EEG)–ECG recordings contain substantial noise, nonlinear temporal patterns, inter‐subject variability, and complex cross‐modal dependencies. To address these challenges, this study proposes a Potter‐Optimized Multiscale Bayesian Gated Axial Quantized Attention Network (PM‐BGAQAN) for robust three‐state seizure prediction using multichannel EEG–ECG signals. The framework first employs adaptive weighted mean filter (AWMF) to suppress physiological noise while preserving relevant signal characteristics, followed by Convolutional Swin‐Transformer Network (CSTN) for hierarchical multiscale spatiotemporal feature extraction and multimodal EEG–ECG representation learning. Inclusion of Bayesian Asymmetric Quantized Neural Network (BAQNN) is followed next for uncertainty‐aware quantized features, whereas multiscale gated axial reverse attention (MGARA) exploits the discriminatory power of multiscale temporal dependencies using gated axial attention. Finally, Potter Optimization Algorithm (POA) optimizes the network parameters to improve classification performance and convergence. The proposed framework is evaluated using the Siena Scalp EEG dataset with timestamp metadata and the SeizeIT2 EEG–ECG wearable dataset for classification of preictal, ictal, and interictal states. Experimental results demonstrate an overall classification accuracy of 99.86%, whereas subject‐independent validation achieves 98.06% average accuracy across the evaluated datasets. Ablation analysis further confirms the contribution of the individual components to the integrated framework. The results indicate that PM‐BGAQAN provides an effective multimodal representation‐learning strategy for robust seizure‐state prediction and establishes a basis for future real‐time, wearable, and clinically validated seizure monitoring systems.

Developmental NeurobiologyVol. 87(1)
Hindustan Institute of Technology and Science (IN)
EEG and Brain-Computer Interfaces
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon