Identification of COVID-19 from Chest CT Scan Using CNN as Feature Extractor and Voting Classifier

COVID-19 was first identified in Wuhan (China) and swiftly spread over the world, resulting in a global pandemic emergency. It has had a profound effect on everyday living, general well-being, and international finance. Rapid diagnosis of susceptible people is critical. There is no precise testing for COVID-19 except for RT-PCR, which is expensive and time-consuming. Recent studies conducted using radiological imaging techniques suggest that such pictures include characteristics of the COVID-19 infection. The implication of machine learning algorithms in conjunction with chest imaging may aid in the accurate detection of this illness and help to overcome the shortage of specialized physicians. This work aims to construct a model for the automated recognition of COVID-19 infection using chest CT scans. To extract features from patient's chest CT scans, a convolutional neural network was used, and Principle Component Analysis was used to decrease computing cost. The proposed model (an ensemble of machine learning classifiers) was created to offer accurate diagnostics by incorporating the five categories (Normal, Mycoplasma pneumonia, Bacterial pneumonia, Viral pneumonia, and COVID-19). The proposed model reached an accuracy of 99.3%, positive predictive value (ppv) of 99.3%, and sensitivity of 99.2 %.