COVID-19 detection from chest X-ray images using CLAHE-YCrCb, LBP, and machine learning algorithms

Area of Science:

• Medical Imaging
• Machine Learning
• Computer-Aided Diagnosis

Background:

• COVID-19 poses a significant global health threat, necessitating rapid and accurate diagnostic tools.
• Existing machine learning models for COVID-19 detection from chest X-rays (CXRs) face limitations including large data requirements, high computational costs, and long runtimes.

Purpose of the Study:

• To develop a computationally inexpensive and efficient method for COVID-19 detection using CXR images.
• To overcome the limitations of existing deep learning models in terms of data, feature size, and computational resources.

Main Methods:

• Employed Contrast Limited Adaptive Histogram Equalization (CLAHE) for CXR image enhancement.
• Utilized the Illumination-Reflectance model to estimate reflectance from chrominance (YCrCb color space).
• Generated feature vectors using normalized local binary patterns from reflectance (Cr) and YCb, followed by classification with algorithms like Naive Bayes.

Main Results:

• The proposed method achieved high accuracy rates of 99.01%, 100%, and 98.46% on three distinct datasets.
• Naive Bayes demonstrated the most robust performance among the tested classification algorithms.
• The approach requires fewer handcrafted features and less computational power compared to state-of-the-art methods.

Conclusions:

• The developed method offers a cost-effective and efficient solution for COVID-19 detection, suitable for resource-limited settings.
• The accessibility of coding materials and datasets encourages further research and development.
• This prototype can aid regions with a shortage of radiologists in improving diagnostic capabilities.