X-Ray classification to detect COVID-19 using ensemble model

Abstract

Diagnosis with X-Rays and other forms of medical images has soared to new heights as an alternative visual Covid infection detector. Radiographic images, primarily CT scans and X-Rays images play massive roles in assisting radiologists to detect and analyse severe medical conditions. Computer-Aided Diagnosis (CAD) systems are used successfully to detect diseases such as tuberculosis, pneumonia and other common diseases from chest X-ray images. CNNs have been widely adopted by many studies and achieved laudible results in the eld of medical image diagno- sis, having attained state-of-art performance by training on labeled data.This paper aims to propose an Ensemble model using a combination of deep CNN architectures, which are Xception, InceptionResnetV2, VGG19, DenseNet-201 and NasNetLarge, that can aid in the diagnosis of various diseases using image processing and arti - cial intelligence algorithms to quickly and accurately identify COVID-19 and other coronary diseases from X-Rays to stop the rapid transmission of the virus. In our experiment, we have used classi ers for the Xception model, VGG19, and Inception- Resnet model. We have compiled a CXR dataset from various open datasets. The compiled dataset was lacking 1000 images for viral pneumonia in comparison with Covid-19 and Normal CXRs, We used image augmentation and focal loss to com- pensate for the unbalanced data and introduce more variation. After implementing the focal loss function, we were able to get better results. Moreover, we implemented transfer learning on these models using ImageNet weights. Finally, we obtained a training accuracy of 92% to 94% across all models. Our Accuracy of the Ensemble Model was 96.25%.