|
|
 |
| |
|
|
|
Diagnosis of COVID-19 from X-rays Using Recurrent Neural Network |
|
|
|
PP: 2279-2284 |
|
|
|
doi:10.18576/isl/110634
|
|
|
|
Author(s) |
|
|
|
S. Hussein,
M. Otair,
S. Alzoubi,
A. Al-Sayyed,
S. Numan,
A. Numan,
S. Jamal,
R. Al-Sayyed,
|
|
|
|
Abstract |
|
|
| Nearly two years ago, the COVID-19 pandemic caused by the SARS-CoV-2 virus has caused drastic changes in many aspects of life at many levels in the world, and this has affected peoples lifestyles. This impact was particularly significant and impactful on the health sectors, among many others. The COVID-19 virus has essentially increased the demand for treatment, diagnosis and testing. The definitive test for diagnosing COVID-19 is reverse transcriptase polymerase chain reaction (RT-PCR); nevertheless, chest x-ray is a quick, effective and inexpensive diagnosis to detect possible pneumonia associated with COVID-19. In this study, the feasibility of using a deep learning-based Recurrent Neural Network (RNN) classifier to detect COVID-19 from CXR images is investigated. The proposed classifier consists of an RNN, trained by a deep learning model. The RNN identifies abnormal images that contain signs of COVID-19. The experiment used in the study employed 286 COVID-19 samples from the Kaggle Repository. The proposed technique is compared with the decision tree algorithm in order to prove the efficiency of the proposed one. The results revealed that the accuracy of the RNN was 97.90%, with a low data loss rate of 2.10%, while the decision tree accuracy was 75.8741%, and a relatively high data loss rate of 24.1259%. These results support the usefulness of the proposed deep learning-based RNN classifier in pre-screening patients for triage and decision-making before RT-PCR data are available.
|
|
|
|
|
|
|
|
