Edge-AI Enabled IoT Framework for Real-Time Stroke Risk Prediction Using Multimodal Biosignals

A stroke happens when there is an obstruction or diminution of cerebral blood flow, culminating in neuronal death, which may end in mortality or significant disability. This study delineates a real-time stroke risk prediction model utilising various biosignals ECG, PPG, blood pressure, oxygen saturation, and body temperature within an Edge-AI enabled IoT framework. The integration of wearable IoT sensors with edge computing guarantees real-time analytics that are energy-efficient and safeguard patient privacy, rendering it suitable for continuous monitoring. In contrast to traditional methods that utilise deep learning, the model presented here employs Temporal Convolutional Networks (TCNs) and Gated Recurrent Units (GRUs) to continually process crucial data streams and convert them into actionable knowledge. Furthermore, to enhance precision in dynamic environments, a lightweight Transformer's long-range signal acquisition module facilitates the transfer of dependencies across signals without utilising feedback loops. The model's assertions concerning prompt and precise stroke risk assessments were experimentally validated using standard biosignal datasets for both testing and training. The model achieved outstanding results, with an accuracy of 99.89 %, precision of 98.4 %, recall of 96.5 %, and a high F1-score, clearly outperforming traditional models including Random Forest, MLP, XGBoost, and CNN-LSTM.