Disease Detection using Deep Learning Algorithms on the Hardware Platforms

Covid-19 virus, which emerged at the end of 2019, brought human life to a standstill all over the world, causing many people to become permanently ill and die. Since its emergence, the health system has come to the point of collapse with its rapid spread all over the world. Despite the uninterrupted work of healthcare professionals and fighting with their whole selves, this virus spreaded rapidly and infected many people in the world and caused death. Covid-19 virus also caused permanent lung damage in some of the people who survived this disease. In this article, an answer is sought to detect the virus that causes Covid-19 disease by using machine learning methods. The aim of the study is to detect the Covid-19 disease quickly and to start the treatment process immediately. In this work, different models were designed using X-Ray images of patients with and without Covid-19 disease, and among these models, the most accurate and fastest result was proposed. In this sense, sample data were produced from existing data by applying Zoom Range, Shear Range and Horizontal Flip data augmentation methods, since data on Covid-19 is not much. In addition, improvements were made using CNN, VGG16, DenseNet121 and ResNet50 deep learning methods to design proposed model. Since the main aim of the study is to achieve the highest accuracy rate quickly, the performances of deep learning algorithms in different working environments were evaluated. CPU, GPU and TPU are used for this. As a result of experimental studies, it has been observed that all algorithms working with GPU work faster with or without data augmentation. In addition, although deep learning algorithms have been successful in working with big data, it has been seen in this study that there is no need for data augmentation for Covid-19 disease detection such a dataset. By examining such image data on the GPU with any deep learning algorithm proposed in this study, we can detect the disease successfully and quickly.