A deep learning approach with Bayesian optimized Kernel support vector machine for Covid-19 diagnosis

The unexpected outbreak of the corona virus in 2019 has had a devastating impact around the world, resulting in unprecedented circumstances. The corona virus spread at an exponential rate, creating a severe threat to human health as well as the world economy. This situation necessitates rapid detection of the virus and necessary healthcare facilities, as the virus can cause severe casualties if left untreated. Machine learning has advanced to the point that it can now solve a wide range of biomedical problems with high precision. The research focuses on using a deep learning mechanism to identify chest X-ray images of Covid-19 and other pneumonia patients in two- and three-class scenarios. The proposed approach employs the GoogLeNet architecture to extract features that are fed into different classifiers. With the Bayesian Optimisation technique, Kernel Support Vector Machineis found to be the most accurate among various classifiers compared. The model showed an overall accuracy of 98.31% for two-class classification between Covid-19 and non-Covid chest X-ray images and 98.60% overall accuracy for three-class classification problem between Covid-19, healthy and viral pneumonia X-ray images. The proposed system outperformed several existing architectures, and it was also tested using smaller datasets so as to ensure robustness.