A DEEP CONVOLUTIONAL NEURAL NETWORK APPLICATION FOR COVID-19 DIAGNOSIS FROM X-RAY IMAGES


Noğay H. S.

3rd International Congress on Engineering Sciences and Multidisciplinary Approaches, İstanbul, Turkey, 10 - 11 February 2022, pp.544-551

  • Publication Type: Conference Paper / Full Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.544-551
  • Kayseri University Affiliated: Yes

Abstract

The COVID-19 epidemic still continues to be on the world agenda, and the number of cases and deaths from this disease continue to increase. Studies on the subject that can help both diagnosis and treatment are constantly increasing and the epidemic is being fought with solutions and practices. In this study, a deep learning application has been implemented that can help accelerate the diagnosis of COVID-19 and pave the way for solutions that will slow down the epidemic. In the study, it was aimed to classify four categories as COVID-19, Lung Opacity (non-COVID lung infection), Normal and Viral-Pneumonia. The designed pre-trained deep convolutional neural network (DCNN) model was trained with Chest X-Ray (CXR) images and tested with 20% randomly selected data. In order to train and test the pre-trained model, some layers were redesigned using the transfer learning approach and made suitable for quadruple classification. The test result with the accuracy rate of 95.75% was compared with the studies made with different DCNN models, and the success of the study was confirmed.