Evaluation of Lung Involvement in Patients with Coronavirus Disease from Chest CT Images Using Multi-objective Self-Adaptive Differential Evolution Approach

Under the global pandemic of COVID-19 over the last year, the use of image processing techniques and the artificial intelligence algorithms to analysis chest X-ray (CXR) images is becoming important. Determining the lung involvement and percentage development of COVID-19 is one of most important requirements for the hospitalization centers. The most studies in this field belong to the articles based on the deep learning methodologies using convolution neural networks, which are usually implemented to facilitate the screening process. Only a few number of studies are about the determining the percentage of lung involvement and development of coronavirus based on CXR images. The lack of comprehensive datasets of CT images with a large amount of samples is one of the most important issues in this field. Determining of lung infection in COVID-19 patients, based on different CXR images in different days, has its own challenges such as different image sizes, illumination density, radiation dose of X-ray and angle of radiation, which makes it impossible to the implement a simple differential filter on two different images. Using an optimization self-adaptive algorithm with differential and multi-objective approach can improve the performance accuracy with a corresponding reduction in computation time.