Penentuan Luas Lesi Pneumonia Covid-19 Menggunakan Arsitektur Segnet pada Citra X-Ray

Abstract

Lung disease is a highly serious public health issue and a leading cause of death worldwide. One of the diseases affecting the lungs is pneumonia. On December 1, 2019, a new variant of pneumonia caused by a novel virus emerged. This virus is a variant of the coronavirus known as Covid-19. The virus has a higher mortality rate compared to general pneumonia. Common symptoms of viral pneumonia include fever, cough, shortness of breath, chest pain, and fatigue, while Covid-19 pneumonia often presents symptoms such as high fever, persistent dry cough, severe shortness of breath, loss of the sense of smell, fatigue, and body aches . Machine learning has recently been frequently employed for disease diagnosis due to the limited number of radiologist experts available to interpret X-ray results. Additionally, reading test results takes time and introduces the possibility of human error. In this study, the author proposes research that will result in a website to assist healthcare professionals in areas with a shortage of experts in identifying the extent of Covid-19 lesion based on X-ray results using the SegNet architecture. A total of 2,912 Covid pneumonia data and 2,912 normal lung data were used, with 2,331 training data, 290 validation data, and 290 testing data, respectively. The application of the SegNet architecture for lung segmentation and lesion segmentation from X-ray images resulted in a mean IoU of 93.96%, an AUC of 99.24%, and a precision of 97.93%.