Deep Learning for COVID-19 Cases-Based XCR and Chest CT Images

Abstract

The ferocity of coronavirus disease (COVID-19) made CT scans urgently needed as a practical and effective alternative. Emergency departments may favor this alternative due it is mobility, accuracy, speed, inexpensive and less radiologist training. Hence, it is required to develop an automated and accurate detection using chest CT scans. Therefore, this study employs a hybrid predictive model. This predictive model comprises a deep convolutional neural network using WekaDeeplearning4j. It extracts chest X-ray and CT imaging features from pneumonia or other lung diseases training cases. Then, it further refines those features with COVID-19 cases to train selected imaging features capable of classifying three cases. They are classified as positive, pneumonia and negative. This study demonstrates capability and effectiveness of the classifier, for detecting positive cases through an experimental performance. This model is tested on a number of reliable and up-to-date datasets such as John Hopkins University, ieee8023 and Kaggle. Trained models of the classifier showed promising detection results in terms of accuracy and speed.