Journal article abstract
Deep Learning Vs. Machine Learning Based Screening of COVID-19 Using Chest X-Ray Images


Research Areas
Currently no objects available

Publication Details
Author list: Norah Abdulraheem Alzahrani, Amaal Ghazi AlOtaibi, Maha M. AlShammari, Lubna Ibrahim Al Asoom, and J. Francis Borgio
Publisher: Medknow Publications
Publication year: 2021
Journal: Saudi Journal of Medicine and Medical Sciences
Journal acronym: SJMMS
Volume number: 9
Issue number: 1
Start page: 101
End page: 101
Number of pages: 1
ISSN: 1658-631X
Web of Science ID:
PubMed ID:
Scopus ID:
eISSN: 2321-4856


Background: COVID-19 pandemic has affected most countries worldwide; the latest recorded number was 14 million confirmed cases. Some studies experiment on using deep learning models to diagnose COVID-19 using medical imaging such as chest X-ray (CXR) images. However, the use of deep learning and machine learning are limited.

Objectives: This study aims to classify the CXR images into either normal or infected by COVID-19 using deep learning and machine learning and comparing the two approaches.

Methods and Material: Our dataset contains 143 images of posteroanterior view CXR for normal (n=67) and COVID-19 cases (n=76), 100 images for training, and 43 for testing. We used the Python programming language to implement the deep learning model to build and evaluate the Convolutional Neural Network (CNN) model. For the machine learning model, we first extract 36 features using Fiji ImageJ tool. We then used these features to build KNN and SVM models in python. We used feature selection and parameter tuning to produce better results for the models.

Results and Discussion: The deep learning model achieved an accuracy of 100% using the CNN model. The best results of the used machine learning techniques were after combining them with feature selection function that achieved 95% for KNN and 98% for SVM. From the results, we found that the deep learning model outperforms the other two machine learning models.

Conclusions: The deep learning model using CNN achieved better results than machine learning for classifying CXR either to COVID-19 cases or normal. Future studies may focus on experimenting using a larger dataset for CXR.


Projects
Currently no objects available

Keywords
Currently no objects available

Documents
Currently no objects available

Last updated on 2021-24-08 at 08:57