An approach for COVID-19 detection using deep convolutional features on chest X-ray images

Zuherman Rustam, Sri Hartini, Ilsya Wirasati, Jane Eva Aurelia

Research output: Contribution to journalArticlepeer-review

Abstract

First screening of COVID-19 becomes very crucial because of its fast spread. There are several ways to diagnose someone who has COVID-19, but chest X-ray is one of the efficient tools that can be used. Deep learning, especially Convolutional Neural Network (CNN), is commonly utilized in medical images due to its superiority in extracting high-level features of images. However, in order to train CNN, we need enormous data to avoid overfitting. Meanwhile, there is a limit of chest X-ray availability that can be access publicly. Considering this problem, we propose pre-trained CNN model as a feature extractor, and the feature vector obtained as the output of CNN that is used as the input of machine learning classifier, namely Support Vector Machines (SVM), Random Forest (RF), and k-Nearest Neighbors (kNN). Using the data from Kaggle COVID-19 Radiography Database, our proposed method with SVM as a classifier succeeded in delivering accuracy of 99.73% in the testing data. Moreover, the performance of CNN-SVM held on training data provides the average accuracy of 99.77%. Thus, our proposed approach can be used as an alternative on screening COVID-19.

Original languageEnglish
Pages (from-to)1452-1460
Number of pages9
JournalJournal of Theoretical and Applied Information Technology
Volume99
Issue number6
Publication statusPublished - 31 Mar 2021

Keywords

  • Convolutional neural network
  • Feature extraction
  • Hybrid method
  • Medical image

Fingerprint

Dive into the research topics of 'An approach for COVID-19 detection using deep convolutional features on chest X-ray images'. Together they form a unique fingerprint.

Cite this