A hybrid approach of Deep Learning Algorithms for Identification of COVID-19 disease using Chest X-Ray Images

Abstract

Machine Learning(ML) algorithms can revolutionize the entire background of the healthcare system and have been used widely to identify diseases efficiently by classifying the image samples. The medical images are created by using visual representations of the interior body parts. ML system computes the distinctive features of the image that are supposed to be very helpful in making decisions regarding the prediction or diagnosis of the disease. It also discovers the best combination of these features that can correctly classify the image. In recent times, deep learning(DL) which is a form of ML mostly being used to analyze medical images for its significant performance in many applications. The DL models can identify the disease from radiology images. In this article some DL algorithms have been used to detect COVID-19 diseased patients from their chest X-ray images(CXR). Three popular convolutional neural networks(CNN) like ResNet, GoogleNet and AlexNet were used to extract features from the dataset. The principal component analysis(PCA) technique was also used which further reduced the dimensions of the dataset. The extracted features were given to the classifiers like SVM and KNN as input in order to identify COVID-19 disease from the images. The proposed method attained an accuracy level of 97.7% with KNN and 98.1% with SVM. The sensitivity and specificity of the model were estimated as 97% and 98% respectively which shows the efficiency of the model for identifying the disease correctly.