Contour Based X-Ray Image Classification System for Detection of Covid-19

Abstract

COVID-19 is a worldwide epidemic in recent times, as announced by the World Health Organization in early 2020. The disease is extremely contagious, affecting the respiratory system and affecting oxygen saturation in the circulatory system. In the current scenario, there are many testing procedures and processes to diagnose COVID-19. These procedures majorly face the challenges of accuracy and delay response because of their testing mechanisms. Machine learning techniques make it easy to support effective diagnostics with images and text data, but it requires a huge amount of test and training data, which requires a huge amount of memory and processing time. In support of efficient quality inference and minimal latency as the main challenge of COVID-19 detection, there are several methods that can provide a better solution to overcome these challenges through the use of cutting-edge trending technologies such as digital imaging techniques. In this study, as a first step to demonstrate machine learning algorithms to achieve high performance, digital imaging techniques were tested through the analysis, classification, and segmentation of chest x-ray (CXR) images. The proposed model starts with feature extraction from input chest x-ray images to determine if the image is COVID-19 or non-COVID-19. This diagnostic model delivers the most interesting results with 96.5% accuracy, minimal latency, and cost-effective standardized COVID-19 tests.