COVID-19 Detection Using Fourier–Bessel Series Expansion-Based Dyadic Decomposition and Custom CNN

Millions of people have died worldwide with COVID-19, which is spreading quickly. A medical investigation has demonstrated that COVID-19 affects patients’ lungs and manifests pneumonia symptoms. For a quick and precise diagnosis of COVID-19, X-ray scans with machine learning may be helpful. The issue of fewer testing kits and physicians can also be resolved by it. In this study, we present a dyadic decomposition approach for images based on the Fourier–Bessel series expansion. X-ray images are divided into sub-band images using this FBD. The modified CNN architecture is then fed to each sub-band picture that was obtained. We have used COVID-19 Radiography Database that includes total of 15,153 X-ray images among which 3616 images are of COVID-19 positive cases. Our suggested model has achieved an accuracy of 99.65%.