Predicting infection with coronavirus wirelessly based on artificial neural network and MATLAB.

Due to the huge fast spread of Covid-19 around the world, which resulted in the loss of many lives, the maximum level of emergency was triggered all over the world. The best way to reduce COVID-19 infection is to prediction it early based on artificial intelligence (AI). To determine whether the patient has COVID-19 infection or not. An accurate and effective diagnosis system for Covid-19 was proposed in this paper. The diagnostic parameters for right and left lungs, D-dimer, and physiological parameters such as SpO2, temperature, and heart rate were collected from CT scans (three RGB colors) for right and left lungs, D-dimer, and physiological parameters such as SpO2, temperature, and heart rate. The data was collected from 300 patients, with each patient receiving 10 samples; 114 of them were infected with Covid-19, while the remaining 186 were uninfected. For training, testing, and verifying the gathered data, an artificial neural network (ANN) on one hidden layer at 20 nodes-based Backpropagation method was used. For all diagnostic parameters, a total of 30,000 samples were obtained (300 patient x 10 parameters x 10 samples per patient). The 3,000 data samples (300 individuals, 10 samples each) were divided into three datasets: 70% for training ANN (2,100 out of 3,000 samples), 15% for testing ANN (450 out of 3,000 samples), and 15% for validation ANN (450 out of 3,000 samples) (450 out of 3,000 samples). In terms of ANN performance, correlation coefficient, error, mean absolute error (MAE), and histogram, the results of COVID-19 diagnosis based on ANN were studied. The MAE for training, validation, and testing at 20 nodes, respectively, was 0.0012, 0.012, and 0.013, indicating that the ANN achieves good diagnostic accuracy. In training, validation, and testing at 20 nodes, the correlation coefficient (R2) between the actual and estimated value was 0.9999, 0.9996, and 0.9998, respectively. In terms of correlation coefficient and MAE, the suggested technique beat the current state-of-the-art. [ABSTRACT FROM AUTHOR]