A hybrid system for detection and diagnosis of novel corona virus

COVID-19 (The Novel Corona virus disease 2019) is an acute respiratory system infection disease caused by a newly discovered corona virus (SARS-CoV-2), emerged in Wuhan, China and spread across the globe since the beginning of 2020 and the greatest crisis of the modern era now. Due to the limited number of test kits accessibility for COVID-19 detection in hospitals as number of cases increasing every day, it is essential to implement an automatic detection system as a quick alternative diagnosis option to prevent COVID-19 spreading among people. In this research twelve fine-tuned convolution neural network based five best models are proposed. AlexNet , GoogleNet, ResNet50, VGG-19, Inception-v3, Inception-Resnet-v2, SqueezeNet, ShuffleNet, Xception, MobileNet-v2, NASNetMobile and DenseNet-201 with different architectures have been trained with different hyper parameters using transfer learning approach due to limited number of chest X-Rays and evaluated to propose best solution for the detection of Corona Virus Disease-19 infected patient using CRXs (“Chest X-Rays”). The result of the experimentations shows that the developed models can provide accuracy with minimal loss for AlexNet (99.84%, 0.0042), GoogleNet (99.84%, 0.0050), ResNet50v (99.68%, 0.0145), VGG-19 (100%, 0.0009), Inception-v3 (98.89%, 0.0311), Inception-Resnet-v2 (99.37, 0.0319), SqueezeNet (98.89%, 0.0253), ShuffleNet (99.37%, 0.0200), Xception (98.89%, 0.0475), MobileNet-v2(99.52%, 0.0200), NASNetMobile ( 98.89%, 0.047) and DenseNet-201(99.37%, 0.0172). In this work we developed a MATLAB app based on these proposed models. This application performs different steps to remove artifact and to segment lungs. At the final step proposed system classify Chest X-Rays either into COVID-19+ or normal.