An efficient IoT-Artificial intelligence-based disease prediction using lightweight CNN in healthcare system

Abstract

In recent years, chronic disease (heart or kidney disease) is a leading cause of death worldwide. Chronic disease prediction is a very complicated task. Doctors who are experienced and familiar with this disease can only be able to predict chronic disease. In a healthcare system, the Internet of Things (IoT) is the key technology. In this paper, an EO optimized Lightweight Automatic modulation classification Network named the EO-LWAMCNet model is proposed to accurately predict a patient's chronic health condition (kidney or heart disease). A sensor implanted in the patient's body can able to collect every data and it uses a gateway to transmit the data toward the cloud. Based on the achieved sensor data, the EO-LWAMCNet model initiates the classification process to predict chronic disease. The model undergoes a testing and training stage. The disease is predicted using CKD and HD datasets. Here, the preprocessed data is used for classification in the training stage. Once the training process is completed, the Cloud server's (CS) sensor data is tested and categorized into abnormal (heart or kidney disease) and normal. The awareness message is sent to the doctor to treat a patient in case of an abnormal result. The performance of the model is evaluated using accuracy, MCC, F1-score, and miss rate. This model can accurately predict the presence or absence of heart or kidney disease with an accuracy of 93.5% using the CKD dataset and an accuracy of 94% using the HD dataset. Also, the miss rate of the model is less in classification.