Dynamic Pulmonary Vessel Analysis in Free-Breathing MRI using 2D-U-Net Segmentation

Motivation: Chronic lung diseases like chronic obstructive pulmonary disease and pulmonary hypertension remodel the pulmonary vasculature, affecting lung function and right heart performance. Goal(s): This study aimed to analyze the central pulmonary vasculature using automated 2D-U-Net segmentation in free-breathing lung MRI, assessing reproducibility and disease-related differences. Approach: Data from healthy participants, COPD-, and PH-patients were used to calculate dynamic vessel parameters, including vessel voxel counts, expansion between systole and diastole and phase shifts using 2D U-Net segmentation. Results: Significant differences were observed in vessel size variability and phase shift between groups, highlighting potential non-invasive markers for distinguishing lung diseases. Impact: This study demonstrates the potential of automated MRI-based vessel analysis to identify distinct hemodynamic changes in chronic lung diseases like COPD and PH. It could offer a non-invasive method to differentiate disease subtypes and monitor treatment responses in clinical practice.