Automated Medical Image Processing for Lung Pneumonia Diagnosis Based on LS-SVM

Abstract

Pneumonia is an inflammation of the lungs that causes pain when breathing and limits oxygen intake. Pneumonia can be caused by bacteria, viruses, and fungi. Image processing, a branch of informatics or computer science, is a field highly related to the manipulation and analysis of digital images. This study aims to design a medical image processing system as an alternative to support the diagnosis of Pneumonia in the lungs using the LS-SVM method. LS-SVM (Least Square Support Vector Machine) is a simpler and modified model of the SVM method. HoG (Histogram of Gradient) is a commonly used feature extraction method in image processing and object detection. The objective of this study is to improve the quality of healthcare services and assist in faster and more accurate clinical decision-making. The results show that lung image analysis using the LS-SVM method has a good accuracy level in the image classification process, with 2000 training data inputs processed in the preprocessing stage, consisting of 1000 Pneumonia images and 1000 normal lung images, while the testing data used consisted of 500 images, with 250 Pneumonia images and 250 normal lung images. Based on the tested data, the system achieved an accuracy of 81% for 1300 tests, proving that the LS-SVM method is effective in image processing with satisfactory results.