Deep Learning Model for Diagnosis of Corona Virus Disease from CT Images
Purpose: SARS-COV-2, a severe acute respiratory syndrome, has caused more than 1 million to be infected worldwide. Corona Virus Disease, known as COVID-19, has cases that are worldwide and widespread. What is worse, this number continues to increase. Early diagnosis of COVID-19 and finding high-risk patients with a worse prognosis for early prevention is vital. It is essential to screen as many as suspect cases for appropriate quarantine and treatment measures to control the spread of the disease. The viral test based on samples taken from the lower respiratory tract is the critical standard of diagnosis. However, the availability and quality of laboratory tests in the infected area may cause inaccurate results, false positive.
Methods: In this study, a Deep Learning (DL) method for COVID-19 diagnostic and prognostic analysis using Computed Tomography (CT)
scans are studied. Based on the COVID-19 CT images, it is aimed to diagnose COVID- 19 at an early stage. Thus, it may take place before a clinical diagnosis before pathogenic testing.
Results: For the testing probability of the disease, 5800 CT images were taken from the Kaggle web. 4640 (80%) CT images are used as the training step, while 1160 (20%) images are benefitted for the testing step. AlexNet has achieved an overall accuracy of 94. 74%, with 87. 37% sensitivity and 87. 45% as specificity while Inception-V4 has an overall accuracy of 84. 14%, with 87. 09% sensitivity, and 84. 14% as specificity. These results show the high value of using Deep learning for early diagnosis of COVID-19. Deep learning is used as a beneficial tool for fast screening COVID-19. In this study, it is benefitted to find potential high-risk patients.
