Fused Deep-Features for Automatic Classification of Chest Radiographs Into Normal/Covid-19

Lung is one of the vital organs which support the respiration and the abnormality in the lung will causes breathing issues. Untreated lung abnormality will case mild to severe health issues. Lung infection due to the Covid-19 is one of the harsh diseases which caused a large infection and death-rate globally. Timely detection and treatment is necessary. This research aims to develop a Deep-Learning-Tool (DLT) to examine the chestradiograph (X-ray) to detect the normal/Covid-19. The proposed scheme includes the stages, like data collection and resizing, feature extraction using ResNet (RN) variants and feature reduction with 50 % dropout, fused-features generation and classification using SoftMax, and performance verification with 5-fold cross validation. This work considered the conventional RN and RNV2 for the investigation and the classification result with DLT is verified using the individual-features (IF) and the fused-features (FF). Compared to the IF, the FF provides better detection accuracy.