Pneumonia Detection Using Enhanced Convolutional Neural Network Model on Chest X-Ray Images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pneumonia Diagnosis

Background:

• Pneumonia is a contagious lung infection requiring early detection and treatment, especially for vulnerable populations like the elderly and young children.
• Untreated pneumonia can lead to severe complications.

Purpose of the Study:

• To develop and compare deep learning (DL) models for detecting pneumonia in chest X-ray images (XRIs).
• To evaluate model performance using accuracy, precision, recall, loss, and ROC AUC scores.

Main Methods:

• Employed DL algorithms including enhanced CNN, VGG-19, ResNet-50, and fine-tuned ResNet-50.
• Trained models on a dataset of 5863 chest XRIs categorized into train, validation, and test sets.

Main Results:

• The enhanced CNN model achieved the highest accuracy at 92.4%.
• ResNet-50 showed the lowest accuracy at 82.8%.
• Developed techniques outperformed popular ensemble methods and cutting-edge approaches.

Conclusions:

• The enhanced CNN model is the most effective for pneumonia detection based on high accuracy.
• DL models, particularly enhanced CNN and ResNet-50, can significantly improve diagnostic accuracy and patient outcomes.