Hybrid Deep Learning and Support Vector Machine for Covid-19 diagnosis based on Chest X-ray images

Document Type : Original Article

Authors

1 Electrical and Bioelectric Engineering Department, Khorasan Institute of Higher Education, Mashhad, Iran

2 Khorasan Institute of Higher Education

Abstract

Under the daily growing of corona disease, the use of artificial intelligence algorithms especially convolutional neural networks (CNN) based on deep learning for screening chest X-ray (CXR) images of Covid-19 patients is becoming more important and encompasses the most articles in the field. CNN detection and screening accuracy in scientific articles has reached more than 95% and the combination of several different CNN architectures has resulted up to 99% accuracy. In this paper, we use a hybrid AlexNet as a convolutional neural network with support vector machine (SVM) as a classifier, to screen Covid-19 infected patients based on chest X-ray images of GitHub database. At the input part of the proposed structure, for feature extraction of images, a two-dimensional histogram which contains more image information in smaller dimensions has been used. The problem of lack of images has been solved by using the data augmentation technique. Simulation results have been compared with other clustering methods such as k-means, fuzzy C-means (FCM) and subtractive clustering. The proposed structure has more than 99 percent sensitivity in diagnosis of corona disease.

Keywords

Main Subjects

Advanced Signal Processing

Articles in Press, Accepted Manuscript
Available Online from 11 July 2022

Files

Share

How to cite

Statistics