Covid-19 detection and classification using artificial intelligence and image processing techniques

Covid-19, as is known, is a serious lung respiratory tract disease that has emerged in recent years and can be fatal. Since its emergence, it has spread in many countries and caused a high number of deaths. Due to the ease of disease spread, its rapid transmission, rapid progression and lethality, early diagnosis has increased the importance of. Many medical studies have been conducted in this field since the disease began to appear. In addition to medical studies, computer-aided disease detection and classification studies for auxiliary diagnostic purposes have also gained momentum. Among computer-aided studies, studies using laboratory data, as well as studies using medical images, are especially important. In this thesis study; It is aimed to classify these images using a mixed data set consisting of images with Covid-19, Normal and Pneumonia features obtained from open sources. For this purpose, a model called Enhanced Swin Transformer Based Few-Shot Learning was developed within the scope of the thesis. The performance of the model was measured with metrics such as accuracy, F1-score, specificity, precision, sensitivity, negative predictive value, positive predictive value, Kappa coefficient, and curve sub-area. According to the determined performance metrics, the comparison of the proposed method with the classical methods found in the literature is also included. The findings obtained show that the proposed model is more successful than other methods. In addition, the proposed model reveals that it can be integrated into clinical applications by detecting Covid-19 with high accuracy even in the case of a small number of images.