Özet:

Millions of people have lost their lives due to the Covid 19 pandemic, and inadequate health systems have been overwhelmed in many countries. Determining the intensive care and ventilation needs of Covid-19 patients and thus making predictions about the prognosis of the disease is crucial in terms of the patient's health status and the effective use of health systems. The Covid-19 chest computed tomography (CT) findings dataset created for this purpose consists of ground-glass opacity (GGO), consolidation, crazy paving pattern (CPP), consolidation and ground glass (GGOC), nodule and ground glass classes (GGON). The approach proposed in this study consists of four steps. The VGG16 model was trained with the chest CT findings dataset in the first step. The most discriminative features obtained in the second step were selected using the BORUTA algorithm. In the third step, the most valuable top 200, 300 and 400 features for each image were obtained by ranking method. In the last step these features were classified with Support Vector Machines and Linear Discriminant Analysis. The overall accuracy obtained for the chest CT findings dataset is 97.02%. This successful result, obtained using the dataset to predict Covid 19 disease prognosis with Deep Learning methods, is a crucial innovation in the classification of chest CT findings in viral types of pneumonia.

Anahtar Kelimeler: