Omni-dimensional dynamic convolution feature coordinate attention network for pneumonia classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Pneumonia poses a significant health risk, especially for vulnerable populations.
• Accurate and efficient pneumonia diagnosis is crucial for effective treatment.
• Current deep learning models for pneumonia classification face challenges with accuracy and parameter efficiency.

Purpose of the Study:

• To propose X-ODFCANet, a novel deep learning network for enhanced pneumonia classification from X-ray images.
• To address limitations of existing methods, including low accuracy and excessive model parameters.
• To improve the diagnostic performance of artificial intelligence in identifying pneumonia.

Main Methods:

• Development of X-ODFCANet incorporating a feature coordination attention module and an omni-dimensional dynamic convolution (ODConv) module.
• Utilizing residual modules for robust feature extraction from X-ray images.
• Feature coordination attention module aggregates spatial information; ODConv module extracts and fuses features across four dimensions.

Main Results:

• X-ODFCANet achieved a 3.77% higher accuracy in pneumonia classification compared to ResNet18.
• The proposed model significantly reduced parameters, approximately 2.5 times fewer than existing methods (4.45M parameters).
• Demonstrated effective improvement in pneumonia classification accuracy and model efficiency.

Conclusions:

• X-ODFCANet offers a promising advancement in AI-driven pneumonia diagnosis using X-ray imaging.
• The novel network architecture effectively balances accuracy and computational efficiency.
• This approach has the potential to improve clinical diagnostic workflows for pneumonia.