An Effective Diagnostic System for Identifying Covid-19 and Pneumonia Diseases Using Machine Learning Algorithms

Abstract

COVID-19, a pandemic disease causes a devastating effect on humans. The heavy loss of human lives triggers us to build an effective automated diagnostic system for identifying the disease with good accuracy. People get affected by this disease are increasing day by day to a greater extend. The availability of test kits is limited in numbers and its performance is not satisfactory. Hence the medical practitioners highly rely on radiological imaging. There is an immediate requirement for processing these enormous images that are generated daily to test the patient is infected with the disease or not. COVID-19 has similar characteristics with its related diseases such as viral pneumonia and bacterial pneumonia. Appropriately diagnosing the classes of diseases based on its severity is highly important in the medical domain. Our proposed system uses machine learning algorithms such Linear Support Vector Machine classifier and Logistic Regression classifier and provides remarkable results. The proposed system has experimented with 250 chest X-Rays from each classes of diseases such as COVID-19, viral pneumonia, bacterial pneumonia, and normal healthy subjects. The results are evaluated with performance measures such as Accuracy, Sensitivity, Specificity, F1-score, Positive Predictive Value (PPV), Negative Predictive Value (NPV), False Positive Rate (FPR), False Negative Rate (FNR), and Error rate. Logistic Regression gives accuracy 93.4%, Sensitivity 94.1%, Specificity 92.9%, False Positive Rate 7.1%, False Negative Rate 5.9%., Positive Predictive Value 91.4%, Negative Predictive Value 95.1% and Error rate 6.6% which is comparatively better than Linear Support Vector Machine classifier that shows Accuracy 91.6%, Sensitivity 93.1%, Specificity 90.5%, False Positive Rate 9.5%, False Negative Rate 6.9%, Positive Predictive Value 87.1% and Negative Predictive Value  95% and Error Rate 8.5%.