IoT-based heart disease risk prediction and categorization for healthcare using an optimized progressive residual attention-based holographic convolutional neural network

IoT-based heart disease risk prediction and categorization for healthcare using an optimized progressive residual attention-based holographic convolutional neural network

연구 분야: Networking

학회: Signal, Image and Video Processing

The rapid growth of smart healthcare monitoring equipment can be attributed to the emergence of portable medical gadgets that possess Internet of Things (IoT) capabilities. Disease prevention is aided by the shift in emphasis from traditional in-person consultations to telemedicine brought about by the incorporation of IoT and deep learning (DL) in healthcare. Nevertheless, there are several limitations with the current methods when it comes to informing patients of their health state based on ongoing changes in their health for prompt healing. This study presents a brand-new smart healthcare monitoring system called Progressive Residual Attention Holographic Convolutional Neural Network (PRAHCNN) based on Holographic Convolutional Neural Network (HCNN) integrated with a Progressive Residual Attention Network (PRAN). By detecting the risk factors, the proposed study aims to give an accurate and timely diagnosis of cardiac illnesses. To assess the physiological data in this case, the suggested model continually gathers input data from wearable sensors and uses a neural network to process it. The risk status for cardiac disease is then divided into four categories by the system: low, medium, high, and extreme. Here, an automated telephone call and/or SMS notification, including the patient's location, is delivered to the patient's relative if the patient's condition is deemed to be at high or extreme risk. According to the experimental results, the suggested method uses the Cleveland database to obtain a better accuracy of 99.8% with a lower error rate of 2%, and the Hungarian dataset to reach a higher accuracy of 99.6% and a lower error rate of 4%.