ABSTRACT
Coronavirus disease (COVID-19) is a universal illness that has been prevalent since December 2019. COVID-19 causes a disease that extends to more serious illnesses than the flu and is formulated from a large group of viruses. COVID-19 has been announced as a global epidemic that has greatly affected the global economy and society. Recent studies have great promise for lung ultrasound (LU) imaging, subjects infected by COVID-19. Extensively, the growth of an impartial, fast, and accurate automated method for evaluating LU images is still in its infancy. The present algorithms provide results of LU detecting COVID-19, are very time consuming, and provide high false rate for early detection and treatment of affected patients. Today, accurate detection of COVID-19 usually takes a long time and is prone to human error. To resolve this problem, Information Gain Feature Selection (IGFS) based on Deep Feature Recursive Neural Network (DFRNN) algorithm is proposed to detect the COVID-19 automatically at an early stage. The LU images are preprocessed using Gaussian filter approach, then quality enhanced by Watershed Segmentation (WS) algorithm, and later trained into IGFS algorithm to detect the finest features of COVID-19 to improve classification performance. Thus, the proposed algorithm detects whether the person is COVID-19 affected or not, from his LU image, in an efficient manner. The proposed experimental results show improved precision, recall, F-measure, and classification performance with low time complexity and less false rate performance, compared to the previous algorithms. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.