![]() | Only 14 pages are availabe for public view |
Abstract ”The worldwide public health has been devastated by the coronavirus (COVID-19) epidemic. A computed tomography (CT) scanning is an excellent method for screening COVID-19. Deep learning can assist physicians in identifying high-risk COVID-19 patients and those with too-small lesions to view in CT scans by separating diseased regions. In order to provide an acceptable evaluation for the severity rate of pneumonia, it is crucial to promptly and reliably separate COVID-19 from the CT image. In this thesis, a robust deep learning-based strategy has been designed for recognising and segmenting a COVID-19 lesion from chest CT images, which would introduce accurate computer-aided decision criteria for physicians about the severity rate. Two main stages have been proposed for detecting COVID-19: First, a deep convolutional neural network (CNN) recognises and classifies COVID-19 from CT images. Secondly, supervised algorithms using a deep U-Net structure are used for the segmentation of COVID-19 regions in a semantic manner. In addition, COVID-19 quantification depends on the pixel density of the infected area to determine the severity rate categories: low, moderate, and critical. The proposed system is trained and evaluated on three different CT datasets for COVID-19, two of which are used to illustrate the system’s segmentation performance and the other to demonstrate the system’s classification ability. The thesis results show that the proposed CNN model can classify images more accurately than 0.99, and the proposed U-Net-based supervised model is better at COVID-19 segmentation when compared to the other deep learning (DL) models in terms of accuracy, dice, and processing time, with an IoU value of over 0.92.” |