Research Article
Fusing Attention and Convolution: A Hybrid Model for Brain Stroke Prediction
@ARTICLE{10.4108/eetsis.7022, author={R. Bhuvanya and T. Kujani and K. Sivakumar}, title={Fusing Attention and Convolution: A Hybrid Model for Brain Stroke Prediction}, journal={EAI Endorsed Transactions on Scalable Information Systems}, volume={12}, number={1}, publisher={EAI}, journal_a={SIS}, year={2025}, month={4}, keywords={AdaGrad, Brain Stroke detection, CNN, Machine Learning, Transformer}, doi={10.4108/eetsis.7022} }- R. Bhuvanya
T. Kujani
K. Sivakumar
Year: 2025
Fusing Attention and Convolution: A Hybrid Model for Brain Stroke Prediction
SIS
EAI
DOI: 10.4108/eetsis.7022
Abstract
INTRODUCTION: A stroke, a sudden interruption of blood flow to the brain, is a leading cause of disability and death. Early diagnosis is paramount for minimizing brain damage and maximizing treatment effectiveness. OBJECTIVES: Traditional diagnostic methods can be time-consuming and have limited Accuracy. METHODS: This study investigates the efficacy of various machine-learning models for stroke prediction. Specifically, it compares established models like K-Nearest Neighbor, Artificial Neural Network, Long Short Term Memory (LSTM), and stacked LSTM with a newly proposed Transformer Convolutional Neural Network (TCNN) architecture, which fuses Transformer and Convolutional neural network (CNN) models. RESULTS: The TCNN demonstrates significant promise, achieving a superior accuracy of 98% when optimized with the AMSGrad optimizer. CONCLUSION: These findings suggest that the TCNN architecture has the potential to revolutionize stroke prediction accuracy compared to existing methods, potentially leading to improved patient outcomes.
