Analyzing CT Scan Images Using Deep Transfer Learning for Patients with Covid-19 Disease

Alodat, M.

The main aim of this research is to improve the experience of medical staff in recognizing three different lung diseases by making an analysis for chest Computed Tomography (CT) scan in the Sultanate of Oman. To facilitate differential diagnosis for patients with respiratory diseases;we used Deep transfer learning (DTL) from pre-trained network on ImageNet of convolutional neural networks (CNN) through using Fine-Tuning on Keras and TensorFlow 2.0 with tf.keras. The first purpose of the research is to Classify chest CT results either positive which means infected patients, with Covid-19, pneumonia viral or pneumonia bacterial. The other outcome is chest CT result is negative, so no-infection. The second purpose is improving the CNN architecture and to overcome its defects. The results of this study revealed that the best performance was chosen among five pre-trained network and it was ResNet50 model, which showed accuracy with (99%). After the chest CT image has been analyzed, we were able to match the actual diagnosis of the seven volunteer patients out of 8 (87.5%) the eighth patient (12.5%) was classified as covid-19 positive but actually the volunteer has no infection. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.