Ensemble Model for COVID-19 detection from chest X-ray Scans using Image Segmentation, Fuzzy Color and Stacking Approaches

摘要

Coronavirus is a virus of RNA-type that can infect both humans and animal and causes a wide variety of respiratory infections. In humans, it also causes pneumonia. Since coronavirus has been declared a pandemic, Reverse Transcription Polymerase Chain Reaction (RT-PCR) has been the standard method for detection but is a time consuming operation and due to sudden surge in demand it has a high cost. In this study, coronavirus was detected from X-ray scans of chest using a deep learning model consisting of fuzzy image enhancement, offline data augmentation, image segmentation and classification through Convolutional Neural Network. For training and classification, an ensembeled model consisting of the features of VGG-16, ResNet-50 and MobileNetV2 was built and optimized with bayesian optimization. The proposed model achieved an overall accuracy of 96.34%. The precision, recall and F1-Score for COVID-19 class was 100%, 96% and 98% respectively.