Lung Radiomics Features Selection for COPD Stage Classification Based on Auto-Metric Graph Neural Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Chronic obstructive pulmonary disease (COPD) is a progressive lung disease requiring effective preclinical health management.
• Accurate staging of COPD is crucial for tailored treatment strategies in this patient population.

Purpose of the Study:

• To develop and validate a novel approach for COPD stage classification using advanced machine learning techniques.
• To enhance the accuracy of COPD diagnosis and staging through the integration of radiomics and graph neural networks.

Main Methods:

• Segmentation of lung parenchyma from high-resolution computed tomography (HRCT) images using ResU-Net.
• Extraction of lung radiomics features using PyRadiomics.
• Construction of a novel lung radiomics combination vector (3 + 106) via generalized linear model (GLM) and Lasso for risk factors and node features.
• COPD stage classification using an auto-metric graph neural network (AMGNN) with a meta-learning strategy.

Main Results:

• The proposed AMGNN model, utilizing the novel radiomics combination vector, achieved superior performance compared to convolutional neural networks and traditional machine learning models.
• Achieved high performance metrics: accuracy (0.943), precision (0.946), recall (0.943), F1-score (0.943), and area under the curve (ACU) (0.984).
• Demonstrated the effectiveness of the developed method for accurate COPD stage classification.

Conclusions:

• The novel lung radiomics combination vector combined with the AMGNN offers a highly effective and accurate method for COPD stage classification.
• This approach holds significant potential for improving preclinical health management and treatment strategies for patients with COPD.
• The study highlights the power of integrating radiomics and deep learning for complex respiratory disease diagnosis.