| International Journal of Applied Information Systems |
| Foundation of Computer Science (FCS), NY, USA |
| Volume 13 - Number 5 |
| Year of Publication: 2026 |
| Authors: Bob Chile-Agada, Laud Charles Ochei, Fubara Egbono |
10.5120/ijais1e51b1d70335
|
Bob Chile-Agada, Laud Charles Ochei, Fubara Egbono . A Recurrent Neural Network-based Architecture for Early Stroke Detection and Diagnosis using Electromyography (EMG) Data. International Journal of Applied Information Systems. 13, 5 ( Nov 2026), 78-85. DOI=10.5120/ijais1e51b1d70335
Stroke remains a leading cause of mortality and long-term disability worldwide and imposes a heavy economic burden on health systems, families and societies. Early detection is critical for timely intervention, yet current diagnostic pathways often depend on expensive imaging, suffer from high false-positive rates and scale poorly in resource-limited settings. This paper presents an integrated Recurrent Neural Network (RNN)-based architectural framework for the early detection and diagnosis of stroke from surface Electromyography (sEMG) data, together with a scenario-based evaluation of that framework. The architecture exploits the sequential learning capability of Gated Recurrent Unit (GRU) and CNN-GRU encoders to model the temporal patterns of multichannel EMG signals, and it combines a robust preprocessing and windowing pipeline with gradient clipping, learning-rate scheduling and class-weighted training to stabilise learning on small, imbalanced cohorts. The framework is evaluated with the Architecture Trade-off Analysis Method (ATAM) across eight quality-attribute scenarios, covering routine clinical screening, class imbalance, high-volume population screening, corrupted signals, offline rural deployment, Electronic Health Record (EHR) integration, model updating and patients with severe motor impairment. The evaluation identifies the sensitivity points, trade-offs, risks and non-risks of the design and the anticipated impact of architectural changes on accuracy, efficiency, scalability, robustness and interpretability. The contribution is an architectural design and its evaluation, not a clinical validation study; a subject-wise validation protocol for future empirical work is specified.