A novel approach for detection of coronavirus disease from computed tomography scan images using the pivot distribution count method

Detecting a person infected with coronavirus disease is a challenging task. In some cases, the infection is asymptomatic in nature. A computed tomography scan of images provides information about lungs and helps detect coronavirus disease infected lungs. An accurate algorithm helps the specialist to detect corona virus-infected lungs easily. In this article, a new technique called the pivot distribution count method has been proposed to extract the texture features of computed tomography scanned images of lungs and apply it for the detection and analysis of coronavirus disease. The technique is compared with a recently developed methodology called pixel range calculation method and some state-of-art methods like local binary pattern and gliding box method. The 'severe acute respiratory syndrome coronavirus-2 computed tomography scan dataset' was used for our experiments. The experimental results show that the pivot distribution count method produces better accuracy for detecting the infection of coronavirus disease with less computational time. It is also observed that the detection accuracy obtained from coronavirus disease infected images is 98% and from non-infected images is 94% using the pivot distribution count method, which is much higher as compared to other methods.