Classification of Lung Disease From Chest X-Ray Images Using Convolutional Neural Networks

The importance of precise diagnosis to contain the spread of the COVID-19 pandemic is highlighted by the worldwide challenge it presents. Alternative diagnostic methods are being urgently sought due to the limited availability of test kits. Recent research emphasizes the usefulness of radiographs, especially CT scans and X-rays, in diagnosing COVID-19 infections. Chest X-rays (CXRs) show quick imaging, are widely accessible, affordable, and portable, which makes them very promising. Utilizing computer-assisted diagnostic methods can improve the precision and dependability of diagnosing COVID-19 through CXR images, diminishing discrepancies among observers and assisting in healthcare decision-making. In this research, we suggest a technique utilizing local phase to enhance the quality of CXR images, ultimately improving their diagnostic usefulness. Our CNN model combines both raw and enhanced chest X-ray (CXR) images as input, incorporating augmented data to enhance diagnostic precision. Ablation studies prove that improved images enhance diagnostic performance. Assessment is carried out on two sets of data: one with 8851 healthy, 6045 pneumonia, and 3323 COVID-19 CXR pictures, and the other for further validation purposes. The results of our model show great promise, with an average accuracy of 95.57% for classifying three categories in Dataset-1. It notably shows outstanding precision, sensitivity, and F1-scores of 99% for detecting COVID-19. In Dataset-2, we achieve a mean accuracy of 94.44%, along with robust precision, recall, and F1-scores for identifying COVID-19. These results highlight how our method can assist medical professionals in quickly and accurately identifying COVID-19 using chest X-ray images. Our approach aids in the continued fight against the COVID-19 pandemic by utilizing advanced computational methods and enhanced data.