Deep Learning Based Stroke Disease Prediction using Multi Modal Biosignals

Stroke is one of the main causes of death and disability in the world. Treatment and diagnosis must begin early in order to improve patient outcomes. Because deep learning is capable of extracting intricate patterns from massive amounts of medical data, it has shown great promise as a tool for predicting stroke illness. Using multi-modal bio-signals, such as electrocardiogram (ECG) and photoplethysmography (PPG), monitored in real-time for the elderly, the proposed system may be able to predict and semantically interpret stroke prognostic symptoms based on deep learning. Designed and implemented an ensemble structure stroke disease prediction system combining CNN, LSTM, and Genetic Algorithm (GA) feature optimization while walking in order to anticipate stroke illness in immediate-time. A dataset of real-time EEG sensor data was used in the development and training of this suggested deep learning-based model for predicting the possibility of a stroke. In light of the actual findings, stroke may be predicted with 98.4% accuracy and 2.8 FPR and 1.5 FNR using the raw data in conjunction with the CNN-LSTM and GA model.