Classification of Lung Diseases Using an Attention-Based Modified DenseNet Model.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Lung diseases pose a significant global health challenge, necessitating accurate and accessible diagnostic tools.
• Computed Tomography (CT) and X-ray imaging are crucial for lung disease diagnosis, with X-rays offering cost-effectiveness and accessibility.
• Early and accurate diagnosis is vital for effective treatment and improved patient outcomes.

Purpose of the Study:

• To develop and evaluate an enhanced deep learning model for improved lung disease detection from X-ray images.
• To leverage Contrast-Limited Adaptive Histogram Equalization (CLAHE) for image enhancement and DenseNet201 for feature extraction.
• To augment the DenseNet201 architecture with hybrid pooling and channel attention for superior performance.

Main Methods:

• Utilized Contrast-Limited Adaptive Histogram Equalization (CLAHE) to enhance X-ray image quality.
• Employed an augmented DenseNet201 model incorporating hybrid pooling and channel attention mechanisms.
• Validated the model's performance against several established pre-trained deep learning models.
• Employed Gradient-weighted Class Activation Mapping (Grad-CAM) for visual interpretability.

Main Results:

• The proposed model achieved high diagnostic accuracy, precision, recall, and F1-scores of 95.34%, 97%, 96%, and 96%, respectively.
• Demonstrated superior performance compared to multiple well-known pre-trained models (VGG16, InceptionV3, ResNet50, etc.).
• Grad-CAM visualizations provided insights into the model's decision-making for identifying normal, pneumothorax, and atelectasis cases.

Conclusions:

• The developed AI model significantly improves the accuracy of lung disease diagnosis from X-ray images.
• The integration of CLAHE and architectural enhancements in DenseNet201 offers a powerful tool for medical image analysis.
• The model's interpretability features can assist radiologists in clinical decision-making and diagnostic processes.