Clustering and Visualizing of Chest X-ray Images for Covid-19 Detection

Abstract

Corona pandemic showed how artificial intelligence has become a part of our daily lives and is breaking into all fields at a high rate and in different ways. Relying on the conventional techniques to test patients such as RT -PCR has two major drawbacks; a long time to get results and a lack of test kits. Therefore, data mining with machine learning techniques has been suggested to investigate covid-19. In this work, chest x-ray image-based covid-19 detection approach is proposed. Three types of x-ray images Covid-19, Pneumonia, and Normal, are used in two frameworks: image visualization and image segmentation. First, the x-ray samples are visualized using histograms to analyze the pixel-value distributions. The visualization approach helps covid-19 specialists to discover the intensity level of infection by examining the corresponding histograms. Second, a segmentation approach is developed with a k-mean algorithm to provide extra image tuning for infected areas. Three different centroids are used to provide different tuning granularity levels. The suggested frameworks give a fast and reliable methodology to help physicians to decide whether there is a virus or not in the x-ray sample. This is done statistically by histograms and visually by monitoring the segmented infected areas.