COVID-19 Classification for Chest X-Ray Images using Deep Learning and Resnet-101

Abstract

The catastrophic spread of the COVID-19 virus has began in December 2019 which first originated in Wuhan, China and spread rapidly throughout the world. The way to break the chain of spread of the virus is by detecting it using a tool called swab and polymerase chain reaction (PCR), but the price of these tools is expensive and the waiting time is long relatively. This study uses Deep learning as an image recognition method with CNN architecture. X-ray images are used as material to identify infected patients with COVID-19 or normal. The total number of x-ray images is 2562 which is divided into 2 classes, positive and normal. The COVID-19 x-ray image will also use CLAHE preprocessing and two sets of data that will be used as deep learning training materials, original data and CLAHE preprocessing data. The training process is conducted using CNN with the Resnet-101 architecture. the experiment divided the data with the ratio of training data and test data of 80:20. The confusion matrix shows the proposed method provides the highest classification performance with 99.61% accuracy, 99.62% sensitivity and 99.60% specificity.