Performance convolutional neural network (CNN) and support vector machine (SVM) on tuberculosis disease classification based on X-ray image

Abstract

Tuberculosis (TB) problem continues to be a significant health issue worldwide, with a specific emphasis on developing countries. Early diagnosis of TB can aid effective treatment and prevent disease transmission. Image-based X-ray classification of tuberculosis has become an interesting research topic in the development of automatic diagnostic systems that can assist doctors in making better decisions. In this research, the performance of the Convolutional Neural Network and Support Vector Machine in the X-ray image-based classification of tuberculosis was compared. The data used consisted of 1400 X-ray images from tuberculosis and normal patients, comprising 700 tuberculosis images and 700 non- tuberculosis images. The results showed that CNN outperformed SVM in the classification of tuberculosis based on X-ray of thorax. CNN achieved an accuracy of 97.86%, while SVM only reached 96.07%. Additionally, CNN also had higher recall and precision values than SVM, indicating that CNN is more suitable for use in X-ray image-based TB classification. In conclusion, this study demonstrated that CNN is superior to SVM in X-ray image-based TB classification. However, further research is needed to enhance the performance of CNN and SVM algorithms in X-ray image-based TB classification using more sophisticated techniques.