Blood Pressure Monitoring and Hypertension Prediction using Dynamic Opposite Learning Based Prairie Dog Optimization Algorithm

Blood Pressure (BP) monitoring and Hypertension (HR) are an essential for preventing and diagnosing Cardiovascular Disease (CVD). However, late diagnosis of BP and HR caused a savior hear disease lead to human death. Photoplethysmography (PPG) signals and Electrocardiography (ECG) are important biological features utilized for diagnosing and detecting hypertension in early stage. ECG and PPG signals suffer from susceptibility and instability due to different medical aspects. This research a Dynamic Opposite Learning based Prairie Dog Optimization with Bidirectional Long Short-Term Memory (DOL-PDO-Bi-LSTM) for monitoring the BP and HR. DOL-PDO utilized to select most relevant features in feature selection to enhance classification accuracy and achieves two common optimization phases such as exploration and exploitation. Preprocessing process utilized min-max normalization technique to remove background noise of images and then feature extraction process utilized EfficientNet model to extract discriminative features. A Bi-LSTM network and multi-scale convolution network is utilized for HR and BP classification accurately. Simulation results illustrates that the proposed DOL-PDO-BiLSTM approach attains better Root Mean Square Error (RMSE) of 2.98, 1.67 and 1.98 on Systolic BP (SBP), Dyostolic BP (DBP) Mean End Expiratory Pressure (MEP) respectively based on the MIMIC-II dataset as compared to existing methods like Arterial Blood Pressure-Net (ABP-Net) and Explainable Hypertension-Net ExHyptNet.