Image Pre-processing techniques comparison: COVID-19 detection through Chest X-Rays via Deep Learning

Abstract

The COVID-19 pandemic has had a very devastating effect and has spread rapidly across the world affecting close to 36 million people. Chest radiography is a very important feature which is used for early diagnosis of various diseases. With the increasing pandemic, there is a growing popularity of training Convolutional Neural Networks (CNN) to diagnose and detect COVID-19 from Chest X-Rays. However, publicly available and medically verified datasets for COVID-19 infected chest X-Rays are scarce, which results in the model not generalizing properly. For this purpose, it is important to pre-process and augment the data being used to train the model. Various pre-processing techniques exist like Global Histogram Equalization (GHE), Contrast Limited Adaptive Histogram Equalization (CLAHE) and Top Bottom Hat Transform. In this review, we study and compare all these pre-processing techniques to understand which is the most suitable for developing a CNN model which can classify an image as being infected with COVID-19 or Viral Pneumonia with high efficacy.