<![CDATA[Pneumonia is a leading cause of child mortality worldwide, and its diagnosis often relies on chest X-ray interpretation, which is prone to human error. This study aims to optimize a Convolutional Neural Network (CNN) model based on transfer learning using the DenseNet-121 architecture for pneumonia classification in chest X-ray images. The model was trained on a Kaggle dataset consisting of two classes: Normal and Pneumonia. Preprocessing included class balancing and data augmentation. Five fine-tuning strategies were tested, ranging from training only the classifier to unfreezing the entire pretrained layers. Evaluation metrics included accuracy, precision, recall, F1-score, and ROC-AUC. Results showed that the strategy of unfreezing Block 3–4 yielded the best performance with 94.39% accuracy, 95.61% F1-score, and 98.04% ROC-AUC. This study demonstrates that selective fine-tuning strategies significantly improve classification performance compared to training only the classifier or the entire network.]]>
Copyrights © 2025
