Accurate Prediction of COVID-19 using Chest X-Ray Images through Deep Feature Learning model with SMOTE and Machine Learning Classifiers

Rahul Kumar; Ridhi Arora; Vipul Bansal; Vinodh J Sahayasheela; Himanshu Buckchash; Javed Imran; Narayanan Narayanan; Ganesh N Pandian; Balasubramanian Raman

We instantiated our proposed methodology with two publicly available datasets: Chest X-Ray Images (Pneumonia) 1 Figure 4 . The number of the images used in the experiment from both the datasets are as depicted in Table 3 and Table 4 . Pre-processing of the Dataset It could be seen that all the acquired X-ray images are of variable shapes and sizes, which increases the difficulty in effective classification. In order to effectively perform classification tasks, image preprocessing is performed. There exist many automatic and semi-automatic techniques for detecting abnormality in the body of the patient, but the absence of reliable and precise techniques can cause ambiguities in the classification process. Keeping aforementioned challenges in the mind, ML based predictive classifiers are used for analyzing the chest X-ray images, which are further discussed in the following sections. X-ray images are taken with a low resolution which may have a variable height to width ratio. In order to facilitate training and testing of the deep networks, necessary pre-processing steps like image cropping and resizing is performed. In the proposed method, all input images are first converted to a standard size of 224 × 224 for a similar course of action in both the developed model.