Implementasi Vision Transformer untuk Klasifikasi Penyakit Pneumonia melalui Citra Chest X-Ray

Abstract

Pneumonia is a type of respiratory infection in the respiratory tract that is often caused by viruses or bacteria. Poor air quality in Jakarta's urban areas increases people's risk of developing pneumonia, acute respiratory infections (ARI), and asthma. 2019 showed that more than 740,180 children under the age of 5 died from pneumonia cases, about 14% of all early childhood deaths. In overcoming pneumonia, medical researchers have conducted many studies related to the problem of early diagnosis of pneumonia. One of the techniques to detect pneumonia is through chest x-rays that have been developed for classification. Vision Transformer (ViT) is one of the Deep Learning architectures developed specifically for image processing. The purpose of this study is to implement the classification task of pneumonia with ViT which is expected to help detect pneumonia early so that it can be treated faster and better. The results of the study show that the ViT model has good performance after applying several variations of augmentation, and is stable both in training and testing. at a small Learning Rate of 0.00001, it produces 80% accuracy for the case of pneumonia disease classification through Chest X-Ray Images.