Deep Belief Neural Network Based Automatic NSTEMI CVD Prediction Using Adaptive Sliding Window Technique

INTRODUCTION: Cardiac Vascular Disease (CVD) is determined to be the most prevailing disease all over the globe specifically in the case of elderly persons. Among various cardiac disease, CVD account for major mortality all over the globe. Diagnosis of cardiac disease at an early stage is mandatory to reduce the rate of mortality. Still, there is no availability of skilled specialists even in case of developed countries for accurate diagnosis. OBJECTIVES: Achieving automated and accurate diagnosis, computer vision based methods that functions with the help of AI techniques are focused on by researchers. In this current research automated CVD prediction model is designed using a deep learning approach. METHODS: ECG image dataset is utilized in this proposed CVD prediction model. Initially, the Non-ST-elevation myocardial infarction (NSTEMI) ECG data collected from the healthcare centre is taken as input. This input ECG image is converted into a signal and further, it is segmented using the sliding window segmentation technique. Then, using segmented signal QRS peak detection is achieved using Elephant Herd Optimization (EHO) algorithm. From the peak, detected signal features are extracted using Heart Rate Variability (HRV) analysis. Following that the extracted features are sent as input into the Deep Belief Network (DBN) classifier to predict CVD patients.  RESULTS: The proposed CVD prediction model is implemented and some of the performance metrics are calculated. Accuracy, error, precision, sensitivity and specificity attained by the proposed model using the second dataset are 95%, 5%. 96%, 94% and 96%. Results showed that the functioning of proposed CVD prediction model is better when compared with other existing techniques. CONCLUSION: Based on this analysis it can be revealed that accurate and timely CVD prediction can be achieved with a lessor error rate. Further, this proposed model can be used in real time healthcare application by collecting NSTEMI ECG signal from patients.