Diagnosis of Covid-19 disease using optimized convolutional neural networks

Document Type : Original Article

Authors

1 Faculty of Electrical and Computer Engineering, Razi University, Kermanshah, Iran.

2 Electrical Department-Electrical and Computer Faculty- Razi University-Kermanshah- Iran

Abstract

The covid-19 pandemic has caused many problems around the world for more than 2 years, so combating with this disease has become one of the important priorities of the human society to overcome these problems. One of the first steps to fight against this disease is the correct and quick diagnosis of this disease, and various methods have been proposed and presented to diagnose this disease. On the other hand, one of the most crucial tools to classification and recognition of images is the use of Convolutional Neural Networks. In this article, using convolutional neural networks, a structure is proposed that receives scanned images of people's lungs and after processing shows the person's status in 3 categories: Covid-19, healthy and pneumonia at the output. One of the most important advantages of this network refers to the high range of accuracy in comparison with other available ways. In addition, the suggested network in this article has been optimized with the use of grey wolf and genetic algorithms. As a result, its final structure has reached a suitable level in terms of complexity, while having much less the number of layers and the number of filters compared to previous networks, and has an acceptable level of accuracy compared to much wider networks. The suggested network has been trained by the Kaggle data set and finally, an accuracy of 95.34% has been obtained for it, which can be achieved higher accuracy if the amount of training data is increased.

Keywords

Main Subjects

Advanced Signal Processing

Articles in Press, Accepted Manuscript
Available Online from 12 November 2022

Files

Share

How to cite

Statistics