CNN-Based Classification of Infectious Lung Diseases using Thorax X-Ray Analysis

Lung diseases are common throughout the world, including chronic obstructive pulmonary disease, asthma, tuberculosis, fibrosis and pneumonia. The risk of death for people with lung disease is 18.7%, meaning that this type of disease needs to be taken seriously. This research presents the development and implementation of deep learning technique to classify lung infections using the Convolutional Neural Network (CNN) method to find the best level of accuracy by making several changes to the number of epochs, models and datasets used. The study utilized a dataset from Kaggle, comprising 1,840 chest x-ray images across four categories: normal, pneumonia, COVID-19, and tuberculosis. The test results for the CNN model had the highest accuracy in the 90:10 data scenario at 96.74%, while the lowest results were in the 70:30 data scenario test at 92.03%. The accuracy value shows that using the 90:10 and epoch 15 scenarios has the most optimal value with a total of 178 correct classifications and 6 incorrect classifications. This study demonstrates the CNN model's effectiveness and practical utility in lung disease classification, suggesting future work to enhance dataset diversity and explore additional deep learning architectures for improved accuracy and broader applications.