Hypertension Identification System for Stroke Patients Based on Photoplethysmography Device

In this paper, a non-invasive portable method for hypertension detection was proposed, using photoplethysmography (PPG) signals and deep learning algorithms. A portable PPG acquisition device was designed based on the MAX30101 photoplethysmography sensor. Data transmission was achieved via a Wi-Fi module. The preprocessed raw data was directly fed into the deep learning model, enabling a deeper extraction of the correlations within the data. The Long Short-Term Memory model (LSTM) is particularly adept at handling long sequence data. It can also avoid the vanishing gradient disappearance and address long-term dependencies. The attention mechanism further enhances the correlation between distantly sampled points, enabling deeper extraction of the PPG signal. The study recruited 9 volunteers (3 with normal blood pressure, 3 with stage 2 hypertension, and 3 with stage 3 hypertension). The results indicate that the proposed model presents good performance, with an accuracy of 0.856, precision of 0.859, recall of 0.856, and an F1 score of 0.857.