Analisa Performa Arsitektur Model Convolutional Neural Network Dengan Variasi Jumlah Hidden Layer Untuk Klasifikasi Tuberculosis Pada Citra X-Ray

Tuberculosis is a deadly infectious disease of the lungs caused by the bacteria Mycobacterium tuberculosis that can be transmitted through the air when a person with tuberculosis coughs, sneezes, or spits and can cause respiratory problems, such as chronic cough and shortness of breath. More than 10 million people are infected every year worldwide, while in Indonesia in 2020, there were more than 390,000 cases of tuberculosis. The diagnosis is often too subjective in detecting tuberculosis, and it is not uncommon for debates to occur between medical personnel or doctors to determine whether a patient is infected with tuberculosis. Therefore, computer vision technology is needed that can detect accurately and quickly. CNN algorithm which is a type of Deep Learning that is widely applied in image classification and can outperform other methods can be used as a method in detecting images. So, in this study, the model analysis and classification of tuberculosis with CNN algorithm using X-ray image data of human lungs were conducted. In this research, the method used is CRIPS-DM and a comparison of 3 CNN models with different number of hidden layers is conducted. Before the computer trains the data, the data is processed first at the data preparation stage which includes re-sizing, gray-scaling, and data augmentation. The results of this research show that the model with 5 hidden layers is the best model that managed to get an accuracy rate of up to 98%. Furthermore, the results of the best model are implemented in a web-based tuberculosis detection application system that can analyse lung X-ray images and output classification results quickly and accurately.