Performance Evaluation of CNN Architectures on NIH Chest X-Ray Dataset with Boosting Ensemble

Lung diseases, including atelectasis, pneumonia, and pneumothorax, are leading causes of mortality worldwide, making early and accurate diagnosis crucial for effective treatment. This study applies a boosting ensemble method for thoracic disease classification using the National Institutes of Health (NIH) Chest X-ray dataset, which contains images across 14 disease categories. Five pre-trained convolutional neural network (CNN) models—VGG16, VGG19(Visual Geometry Group), ResNet50 (Residual Networks), MobileNetV2, and DenseNet201—were fine-tuned to extract features from the chest X-ray images. These models are widely used in medical image analysis due to their ability to automatically extract relevant features from complex images. To address data imbalance, often problematic in medical datasets, a downsampling approach for the majority class is applied. The study further employs a boosting ensemble method to improve classification accuracy of 92.58%. Comparisons between the individual CNN models and the ensemble method show that the ensemble consistently outperforms individual models, especially in handling imbalanced data. This approach demonstrates the potential of ensemble learning in enhancing the accuracy of lung disease detection through medical imaging.