Detection of COVID-19 in Computed Tomography (CT) Scan Images using Deep Learning

Abdulrazak Yahya, Saleh and Ilango, Letchmikanthan (2020) Detection of COVID-19 in Computed Tomography (CT) Scan Images using Deep Learning. International Journal of Advanced Trends in Computer Science and Engineering, 9 (5). pp. 7441-7450. ISSN 2278-3091

PDF Detection of COVID-19 in Computed Tomography (CT) Scan Images using Deep Learning_pdf.pdf Download (23kB)

Abstract

Coronavirus disease (COVID-19) isa fresh genus found in 2019 that was not previously known in humans. On the other hand, Deep learning is one of the most important fields of medical imaging science at present.In this article, a model of deep learning is being trained for the COVID-19 detection in CT Scan images. This study is implemented using Python programming language. To build and train the Convolution Neural Network (CNN) model, Python Deep Learning libraries such as Keras and TensorFlow 2.0 have been utilized. As for the dataset, the open source dataset of COVID-19 chest computed tomography (CT) images were used. These image where been confirmed by the senior radiologist who performed Diagnosis of and treatment of patients with COVID-19. There were total of 745 images belonging to two classes were sampled. 348 positive (+) COVID-19 images and 397 negative (-) COVID-19 images. Based on the training process, the model was able to detect 79 per cent accuracy on the test set. The performance of the model, Convolution Neural Network were evaluated by comparing with Logistic Regression model. Findings from the research proves that Convolution Neural Network are reliable by producing higher accuracy rate of 79% while Logistic Regression produce a rate of 54%. However, in the future more reliable and quality image datasets should be used along with the metadata of the patients to train the model.

Actions (For repository members only: login required)

View Item