Coronavirus infected lung CT scan image segmentation using Deep Learning

Asipong, K.; Gabbualoy, S.; Phasukkit, P.

This research presents an application of lung segmentation of an internal organ from computed tomography scan images using an artificial intelligence development approach. The deep convolutional neural network technique was used to perform semantic segmentation of an internal organ by training our framework through different computed tomography scan image slices with a medical dataset that has abnormal physiology. Coronavirus disease infected on the lung datasets were used as a study case in this research with adjusted deep fully convolutional neural network compared with an architecture of U-Net model that was used to implement the experimental process. However, the this research will aim to develop a deep learning model as an image processing to use with medical image data to reduce the time in a part of treatment planning. The result of this experiment has shown that an adjusted model of 1 layer U-Net shape can achieve a lung segmentation with 92.46% accuracy and less model training time compared to the original U-Net up to 84.76%. © 2021 IEEE.