Pneumonia Image Classification: A Comparative Study of Deep Learning Mode

Pneumonia is an inflammatory lung disease, also a significant global health concern. Therefore, early and accurate diagnosis is critical for improving patient prognosis and reducing mortality rate. Nowadays, the integration of deep learning methods with medical imaging has shown promising results in enhancing pneumonia detection. This study aims to evaluate the effectiveness of the modified CPNet model in pneumonia diagnosis and analyze its classification performance advantages through comparison with three traditional models, including convolutional neural network (CNN), ResNet-50, and Vision Transformer (ViT). For ResNet-50 and ViT, classifiers are Logistic Regression and Random Forest, respectively, which have demonstrated strong performance in previous research, were selected. The experiments used a chest X-ray image dataset to assess the weighted classification performance of these models. The results demonstrate that CPNet outperforms the other models across Accuracy (0.909), Precision (0.916), Recall (0.909) and F1-score (0.906), highlighting its effectiveness as a reliable tool for pneumonia detection.