Severity and Progression Quantification of COVID-19 in CT Images: a new Deep-Learning Approach

The global spread of the COVID-19 has increased the need for physicians and accurate and efficient diagnostic tools. Therefore, artificial intelligence has a good potential to reduce the workload of physicians and diagnose the disease more accurately. The purpose of this article is to evaluate the existing computer systems and finally to present a more efficient computer system for CT scan image segmentation. To address these challenges, a novel artificial intelligence machine learning-based COVID-19 infection segmentation is proposed to automatically identify infected regions and also severity of them in lung CT images. In deep learning (DL) methods it is necessary to solve the following problems: reduce the requirements for computing resources by reducing the dimension of the initial data; and eliminate the essential dependence of the applied classifier on the observed plot. The essence of the proposed approach for solving these problems is to use statistical parameters calculated in blocks of dimension m × m into which the original image is divided as input data for the classifier. The proposed DL-based segmentation system yielded accuracy of 0.97 and precision of 0.97.