Multidimensional cortical morphological alterations in COPD using explainable machine learning

Area of Science:

• Neuroscience
• Pulmonology
• Medical Imaging

Background:

• Cognitive dysfunction is a frequent extrapulmonary complication in chronic obstructive pulmonary disease (COPD).
• Understanding the neuroanatomical basis of cognitive impairment in COPD is crucial for developing targeted interventions.
• Structural brain alterations may underlie the observed cognitive deficits in COPD patients.

Purpose of the Study:

• To identify specific cortical morphological features associated with cognitive performance in patients with COPD.
• To investigate the relationship between lung function, brain structure, and cognitive status in COPD.
• To apply advanced machine learning techniques for analyzing complex neuroimaging data in COPD.

Main Methods:

• Utilized an XGBoost-SHAP machine learning framework to analyze structural MRI data.
• Integrated pulmonary function tests, cognitive assessments, and detailed cortical morphometry.
• Compared neuroimaging and clinical data from 72 COPD patients and 68 healthy controls.

Main Results:

• Multidimensional cortical features effectively differentiated COPD patients from healthy controls.
• Key features included cortical thickness in the parahippocampal and precentral gyri, and local gyrification index in the insula, middle frontal sulcus, and subcallosal area.
• Right precentral cortical thickness showed a significant association with both pulmonary function (FEV1/FVC) and cognitive performance, mediating 29.19% of this relationship.

Conclusions:

• Cortical morphology is significantly related to cognitive function in individuals with COPD.
• Findings suggest a structural connection within the lung-brain-cognition axis.
• The precentral gyrus emerges as a critical region of interest for understanding COPD-related cognitive dysfunction.