Covid-19 Detection Based on Deep Learning Feature Extraction and AdaBoost Ensemble Classifier

Abstract

In January 2020 the World Health Organization (WHO) declared the deadly disease Corona Virus 2 (SARS-CoV-2) or (COVID-19) as a global pandemic. The adopted benchmark test results for the detection of COVID-19 is Reverse Transcription Polymerase Chain Reaction (RT- PCR). The test is time consuming and expensive as well. With the nature of the virus, a rapid and efficient way of testing is needed. With the application of medical imaging in different fields of medicine and with the success of Artificial models in many fields of medicine, COVID-19 detection using Computed Tomography (CT) scan images can serve as an alternative to the RT-PCR test, as CT scan images are used in profiling COVID-19 patients in hospitals. In this study, two types of training were performed on three different pre-trained deep learning models namely ResNet-50, ResNet101, and VGG16 by employing the Transfer learning method, in the first training feature extraction and classification were carried out by the pre-trained models, while in the second training, features were extracted utilizing the pre-trained models feature extraction part and AdaBoost ensemble classifier for classification. ResNet-50 with the AdaBoost ensemble classifier outperformed the state of the art models employed on the same dataset.