Deep Net Model for Detection of Covid-19 using Radiographs based on ROC Analysis
Dhaya R.
Abstract
There is a rapid spread of the novel corona virus (Covid-19) among millions of people and causing the death of hundreds of thousands of people according to the analytical data provided by the European Centre for Disease Prevention and Control. However, the number of test kits available for Covid-19 is still limited despite the continuously increasing cases every day. Implementation of an automatic detection system is essential for diagnosis and prevention of the spread of Covid-19. Chest X-ray radiographs are used for the detection of Corona Virus using three significant models of convolution neural network namely Inception- ResNetV2, InceptionV3 and ResNet50. Among the existing systems, the highest performance and classification accuracy is provided by the ResNet50 model. A novel framework based on CNN model is proposed that offers improved specificity, sensitivity and accuracy when compared to the existing models. Fivefold cross validation is used for analysis of the existing models and comparison of the proposed model by means of confusion matrices and ROC analysis.