Deep Learning Based Covid-19 Detection With A Novel CT Images Dataset: EFSCH-19

Abstract

Listen

Translate article

COVID-19 pandemic has negatively affected the whole world in many ways. Since its inception, various methods and approaches have been developed. The common feature of these solution searches is minimizing the social and economic damages of the COVID-19 pandemic. In this article, we developed our deep learning-based model for the detection of COVID-19 disease from chest CT images. However, we did not use the publicly available datasets used in most studies in the literature. Because, in public datasets; there are problems such as low samples, incorrectly labeled images and unbalanced distribution. Due to such problems, we thought that our model would not reach the desired high accuracy values. We used our dataset, which has not been included in any deep learning study before, from Elazig Fethi Sekin City Hospital, for the first time in the training of our model. Our model was trained with 800 positive and 800 normal chest CT images and then tested with 400 randomly selected test images. As a result of these tests, accuracy rate of %97.5 was achieved. When the results of our study are evaluated, it is thought that it can help physicians in the detection of COVID-19 disease.