Covid19 Identification using Machine Learning Classifiers with Histogram of Luminance Chroma Features of Chest X-ray images

Abstract

The outbreak of the novel Coronavirus has caused catastrophic consequences on the entire global economy leading to a huge loss of health and wealth. Mankind has suffered a lot due to this pandemic. Large number of screening tests are performed on the suspected individuals by using Covid-19 test kits. As the rate of spread of this disease is increasing exponentially, medical organizations are finding it difficult to screen the suspected cases due to limited availability of test kits. Early diagnosis of coronavirus infection can be made from chest X-ray images of an individual. Current paper proposes a color space based global texture feature extraction method to identify covid19 infected cases. Luminance Chroma features of chest X-ray images are extracted from YCrCb, Kekre-LUV, and CIE-LUV color spaces. These extracted features are used for training different machine learning classifiers and ensembles to perform 3-class classification as covid19, pneumonia, and normal. Results computed at 10-fold cross-validation show that ensembles perform better than the individual machine learning (ML) classifiers. Performance of the proposed method is calibrated on an open-source dataset: Covid19 by considering Accuracy, Positive predicted value (PPV), Sensitivity (Recall), F Measure, and Matthew’s correlation coefficient (MCC) performance measures.