Bayesian-Enhanced LSTM-GRU Hybrid Model for Cloud-Based Stroke Detection and Early Intervention

Abstract

Stroke is a leading cause of mortality and long-term
disability, necessitating early detection and timely
intervention to improve patient outcomes.
Traditional stroke monitoring systems often face
challenges related to real-time processing,
imbalanced data, and computational efficiency.
Existing stroke detection systems face challenges in
real-time processing, leading to delayed diagnosis
and intervention. Traditional models struggle with
imbalanced and noisy physiological data, reducing
classification accuracy. Additionally, optimizing
deep learning models for cloud-based deployment
remains complex due to high computational costs.
These limitations necessitate an efficient, adaptive,
and scalable solution for accurate stroke monitoring
and early intervention. This study proposes a
Bayesian-Enhanced LSTM-GRU Hybrid Model for
cloud-based stroke detection and early intervention.
The model leverages Long Short-Term Memory
(LSTM) networks to capture long-term
dependencies and Gated Recurrent Units (GRU) for
computational efficiency. To further enhance
performance, Bayesian Optimization is employed
for hyperparameter tuning, ensuring optimal
accuracy while minimizing resource consumption.
The system integrates cloud computing for real-time
data processing, enabling continuous monitoring
and automated alerts for healthcare providers.
Experimental results demonstrate superior
classification performance, achieving 96%
accuracy, 95% precision, 94% recall, and an AUCROC
of 1.00, with significantly reduced latency
compared to traditional models. The proposed
approach enhances stroke detection reliability,
reduces false positives, and supports proactive
medical intervention, making it a scalable and
efficient solution for cloud-based healthcare
systems.