Abstract

Radiographs play an important role in medical diagnosis. During the Covid-19 pandemic, the role of radiography, especially taking 2D X-ray images, is very important for the diagnosis of Covid-19 patients and also monitoring the progress of the Covid-19 patient's condition based on the patient's chest X-ray image. In the process of diagnosis, the determination of the condition of the lungs based on X-ray images is carried out by visual observation by a radiologist. This is quite subjective because it will be affected by many variables such as experience, the doctor's physical condition, to the human error factor. Therefore, it is necessary to develop a more effective segmentation system. In this study, the UNet-UBNet image segmentation method was developed which focuses on segmenting images of Covid-19 patients. Different input variations are used to determine the effect of input size on the segmentation process. The UNet-UBNet model is also applied to periodically monitor the development of the lung condition of Covid-19 patients. An interesting pattern was found for the association of lung size with the patient's health condition. In this research, a desktop GUI is also developed which functions to segment X-ray images more effectively.

Keywords: Image segmentation, Covid-19, UNet-UBNet.

Presentation

Demo

Demo : Using The Software to Diagnose Covid-19 Patients