Optimized deep learning models for stress-based stroke prediction from EEG signals

In this research, the use of non-invasive electrode-based EEG (electroencephalogram) signal measurement as a means to detect emotional and stress-related factors in individuals, is explored. This research focuses on evaluating EEG signals, particularly those channels that provide insights into motor-related brain activity, cognitive processes, and visual perception and processing. These have attracted interest in the context of predicting diseases. By integrating stress and disease prediction into a single module, this approach aims to offer an early warning system for individuals, enabling them to mitigate or avoid future health risks. Chronic stress is known to have hazardous effects on health, potentially triggering inflammatory responses within the body. These responses are linked to an elevated risk of cardiovascular diseases, which, in turn, can heighten the risk of stroke. The proposed research aims to classify stress-induced emotions and predict stroke risk using advanced deep learning algorithms. The study utilizes EEG signals to categorize stress-related emotions, subsequently assessing stroke risk via an optimized deep learning model. The proposed model is distinguished by its optimized hybrid optimization technique for feature extraction, aimed at stress and stroke prediction. The classification of stress emotions is achieved through the application of a BiLSTM (Bidirectional Long Short-Term Memory) network, while the assessment of stroke risk is conducted using deep Q-learning. The effectiveness of the proposed model is validated through experiments conducted with the benched mark DEAP dataset, demonstrating its robust performance in both stress and stroke prediction.