Effective detection of COVID-19 outbreak in chest X-Rays using fusionnet model

Abstract

ABSTRACT The worldwide pandemic of coronavirus disease 2019 (COVID-19) triggers acute respiratory difficulties and breathing disorders in humans. The reverse transcription-polymerasssse chain reaction (RT–PCR) examination is generally done for the primary diagnosis of COVID-19. Though, it realizes a high false-alarm rate because the test is done during the first five days of disclosure. But still, the medical field hugely relies on the skill of healthcare professionals to interpret the results. Hence, an automatic chest X-ray-based COVID-19 detection system is more essential in recent times. To fulfil this requirement, a novel FusionNet Model is proposed in this paper for the precise classification of COVID-19 diseases. The CXR (Chest X-Ray) images are initially gathered from Cohen CXR Dataset and COVID-CT-MD dataset. The collected data is pre-processed through image resizing, ROI (Region of Interest) extraction, and noise filtering through Improved Anisotropic Filtering (IAF) to attain an enhanced image. The feature extraction process extracts the global and local features by adopting the Parallel Attention Layer (PAL) module. The optimal features are selected based on the Entropy Correlation score – ESA (Emperor Salp Algorithm) based feature selection process. Finally, the selected features are fed into the SoftMax classifier for COVID-19 detection from chest X-ray images. The implementation of the proposed model has been executed on Cohen CXR Dataset and COVID-CT-MD dataset using the Python platform. The classification accuracy of 99.82% is attained in the case of the Cohen CXR Dataset, and 99.02%. In the case of the COVID-CT-MD dataset is obtained in the proposed FusionNet model.