COVID19 X-Ray Image Classification using Voting Ensemble CNNs Transfer Learning

Resumen

COVID-19 is a novel pandemic and infected COVID-19 people are overgrowing, involving an outbreak that is changing lifestyles around the world. A global issue when trying to contain the propagation of the illness is how to efficiently detect infected people and isolate them. Medical image classification is one of the medical screening tools being used nowadays. Apart from manual inspection, several automatic methods can be applied to exploit artificial intelligence, such as Convolutional Neural Networks (CNNs).Transfer learning can also be applied to make predictive models more effective worldwide, due to the expected small amount ofCOVID-19 chest X-ray images available. In this paper, we propose the Voting Ensemble CNNs Transfer Learning to recognize the COVID-19 footprint and classify it in a chest Xray image. The dataset used for training and evaluation was collected from several sources: COVID-19 image data collection and NIH Chest X-ray Dataset of 14 Common Thorax Disease Categories. Our voting ensemble comprises CNNs architectures: ResNet18, ResNet34, and AlexNet. The results illustrate that our model performs with an accuracy of 0.9, recall of 0.825, precision of 0.971, and F1 score of 0.892.