IDENTIFICATION OF COVID-19 USING AI TOOL

Abstract

COVID disease still spreads all over the world through changing its structure and developing a new variant. This new variant of the virus developed through modification of its structure causes worse symptoms in patients than the previous variant. Many scientists worldwide are tracking the structural modification in viruses and looking forward to achieving faster and more automated prediction techniques to deduce the spread of infection. However, the existing prediction techniques face difficulties in attaining early and accurate predictions of infected persons. So, in this current research automated COVID prediction model is proposed to achieve effective prediction. The chest X-ray images collected from medical IoT devices are initially taken as input. The input image is pre-processed using anisotropic diffusion filtering and adaptive gamma correction for achieving noise removal and enhancing contrast. Then, the required region for prediction is segmented from the image using the watershed segmentation technique. At last, COVID infected persons are predicted using a hybrid CNN-LSTM model. The simulation analysis of the proposed automated prediction is done by estimating certain metrics such as accuracy, sensitivity, error precision and specificity. The accuracy, precision, specificity and error value reached for the proposed model is 97%, 96%, 94% and 3%. This analysis reveals that automated early and accurate prediction of COVID disease is achieved by means of the proposed CNN-LSTM model.

Keyword: COVID disease; Medical IoT devices; chest x-ray images Automated prediction, deep    learning techniques. hybrid CNN-LSTM model