Improving Tuberculosis Detection in Chest X-ray Images through Transfer Learning and Deep Learning: A Comparative Study of CNN Architectures

Abstract

Introduction: Tuberculosis remains a significant global health challenge, necessitating more efficient and accurate diagnostic methods. Methods: This study evaluates the performance of various convolutional neural network (CNN) architectures VGG16, VGG19, ResNet50, ResNet101, ResNet152, and Inception-ResNet-V2 in classifying chest X-ray (CXR) images as either normal or TB positive. The dataset comprised 4,200 CXR images, with 700 labeled as TB-positive and 3,500 as normal. We also examined the impact of data augmentation on model performance and analyzed the training times and the number of parameters for each architecture. Results: Our results showed that VGG16 outperformed the other models across all evaluation metrics, achieving an accuracy of 99.4%, precision of 97.9%, recall of 98.6%, F1-score of 98.3%, and AUC-ROC of 98.25%. Surprisingly, data augmentation did not improve performance, suggesting that the original dataset's diversity was sufficient. Furthermore, models with large numbers of parameters, such as ResNet152 and Inception-ResNet-V2, required longer training times without yielding proportionally better performance. Discussion: These findings highlight the importance of selecting the appropriate model architecture based on task-specific requirements. While more complex models with larger parameter counts may seem advantageous, they do not necessarily offer superior performance and often come with increased computational costs. Conclusion: The study demonstrates the potential of simpler models such as VGG16 to effectively diagnose TB from CXR images, providing a balance between performance and computational efficiency. This insight can guide future research and practical implementations in medical image classification.