MV-MFF: Multi-View Multi-Feature Fusion Model for Pneumonia Classification

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Pneumonia is a leading cause of death globally, requiring expert interpretation of chest X-rays for diagnosis.
• Lateral X-rays offer valuable diagnostic information but are underutilized.
• Multi-view imaging enhances diagnostic precision for various diseases.

Purpose of the Study:

• To develop and evaluate a novel artificial intelligence model for pneumonia classification using multi-view X-ray images.
• To improve the accuracy and efficiency of pneumonia diagnosis.

Main Methods:

• Proposed a multi-view multi-feature fusion (MV-MFF) model integrating latent representations from variational autoencoder (VAE) and beta-variational autoencoder (β-VAE).
• Utilized multi-view X-ray data for pneumonia classification.
• Employed deep learning techniques for feature extraction and data fusion.

Main Results:

• The MV-MFF model achieved an accuracy of 80.4% in classifying pneumonia.
• The model demonstrated an area under the curve (AUC) of 0.775.
• Outperformed existing state-of-the-art methods in pneumonia detection.

Conclusions:

• Multi-view X-ray analysis combined with advanced AI models like MV-MFF significantly enhances pneumonia diagnosis.
• The proposed model shows promise for improving clinical decision-making in respiratory disease management.
• Increased utilization of lateral X-rays within AI frameworks can advance diagnostic capabilities.