COVID-19 Diagnosis based on Chest X-ray using Deep Convolution Neural Network and Testing the Software Complexity using Halstead Metrics and Artificial Neural Network
Keywords:
COVID-19, Chest X-ray images, CNN, Transfer Learning, Software engineer, Halstead, ANN.Abstract
The emergence and spread of the COVID-19 virus has cast a shadow over the world and brought normal life practices to an almost complete halt. Among the methods that are used to detect infection with the Coronavirus are pathological tests, computerized tomography (CT) and chest X-ray (CXR) imaging technology.
This study consisted of three stages: first, used the deep learning to diagnosing infection with the COVID-19 by building a convolutional neural network (CNN) from scratch to classify the CXR images. Then, the processed data was classified to COVID-19 or Normal, and used a transfer learning model approach in order to augment the performance of the COVID-19 detection by obtained a deep features. The Visual Geometry Group (VGG16) architecture is used. The model is achieved a train accuracy 99.992% and validation accuracy 94.33%. The best performance of CNN model built from scratch had an train accuracy 98.67% and validation accuracy 94.67%. Second, the software complexity was measured using Halstead complexity metrics. Finally, an artificial neural network (ANN) can be used to estimate software complexity as an alternative method for Halstead metrics. The proposed model can be applied to a program of any length to estimate its complexity.