Motivation: 129Xe MRI/MRS can assess distinct aspects of pulmonary gas exchange and hemodynamics. However, there is no gold standard against which these metrics can be validated. Goal(s): To evaluate whether unsupervised cluster analysis of 129Xe MRI/MRS metrics naturally reveal patterns known to be associated with certain disease groups. Approach: Eight 129Xe MRI/MRS features were subjected to k-means clustering with internal validation indices used to determine optimal cluster number. Results: The analysis identified four clusters that largely distinguished healthy, COPD, and ILD patient groups. Impact: This study offers a pathway for designing future prospective clinical trials that could validate non-invasive 129Xe MRI/MRS metrics of gas exchange by demonstrating that certain patterns distinguish between lung disease subtypes with high accuracy.
Li et al. (Tue,) studied this question.