COVID-19 Detection from Chest X-Rays and CT Scans using Dilated Convolutional Neural Networks
10th IEEE International Conference on Communication Systems and Network Technologies, CSNT 2021
; : 369-374, 2021.
Article
in English
| Scopus | ID: covidwho-1704647
ABSTRACT
WHO has declared “Coronavirus disease 2019” (COVID-19), which is caused by “Severe Acute Respiratory Syndrome Coronavirus 2” (SARS-CoV-2), a worldwide pandemic in March 2020. With its advent in December 2019, it has affected over 86,095,614 people worldwide as of January 4, 2021. Medical workers and researchers are working towards developing a vaccine and improving diagnostic methods for early detection and disease progression monitoring methods. The objective of this study is to provide a robust “Convolutional Neural Network” (CNN) architecture for COVID-19 detection using “Chest X-Rays” (CXR) and Chest CT Scans in order to reduce the response time to diagnose infected patients. We developed deep learning image classification models using Dilated Convolutional Neural Networks as the backend for our model and utilized various fine-tuned pre-trained CNN models as the feature extractor for our model. For both Chest X-Ray and Chest CT, we created datasets by combining various publicly available databases. The Chest X-Ray dataset contains 196 COVID positive frontal CXR images and 196 normal images, and out of two Chest CT datasets, one contains 349 covid and 349 non-covid images and the other contains 1252 covid and 1230 non-covid images. We also utilized transfer learning because of less publicly available data. Image Enhancement Techniques were also used to improve image contrast. The best classification accuracy achieved on Chest X-Ray dataset is 100% and accuracies achieved on the two Chest CT datasets are 91.6% and 98% respectively. © 2021 IEEE.
Full text:
Available
Collection:
Databases of international organizations
Database:
Scopus
Language:
English
Journal:
10th IEEE International Conference on Communication Systems and Network Technologies, CSNT 2021
Year:
2021
Document Type:
Article
Similar
MEDLINE
...
LILACS
LIS