Differentiating Emphysema From Emphysema-Dominated COPD Patients with CT Imaging Feature and Machine Learning

Area of Science:

• Pulmonary Medicine
• Radiology
• Artificial Intelligence

Background:

• Distinguishing emphysema from emphysema-dominant COPD is clinically significant but challenging.
• Quantitative computed tomography (QCT) shows promise for improved characterization.
• Optimal use of QCT with machine learning for this differentiation requires further investigation.

Purpose of the Study:

• To develop and validate a machine learning model using QCT features to differentiate emphysema from emphysema-dominant COPD.
• To explore the relationship between QCT parameters and pulmonary function tests in these conditions.

Main Methods:

• A prospective study included 476 participants (99 emphysema, 377 emphysema-dominant COPD).
• Participants underwent spirometry and chest CT scans.
• A random forest model was trained on QCT features (emphysema index, lung density, airway/vessel measurements) to classify groups.

Main Results:

• The machine learning model achieved high accuracy (AUC-ROC = 0.97) in differentiating the two groups.
• Emphysema index and airway wall thickness were key features for classification.
• QCT-derived emphysema index correlated negatively with FEV1/FVC in emphysema-dominant COPD but not in emphysema alone.

Conclusions:

• Machine learning analysis of QCT features effectively differentiates emphysema from emphysema-dominant COPD.
• Distinct relationships between QCT parameters and lung function suggest different pathophysiological processes.
• Findings support improved diagnostic and management strategies for emphysema and COPD.