Impact of Segmentation Masks on Deformable Image Registration in Post-Processing of Functional Lung MRI

Motivation: Accurate image registration is crucial for functional lung MRI post-processing, enabling calculation of ventilation between respiratory states. Lung segmentation can support this step, though accurately labeling the lung periphery is challenging and may cause inconsistencies between inspiration and expiration masks, impacting registration. Goal(s): To assess the influence of weighting masks for deformable registration on functional ventilation estimation in lung MRI. Approach: Lung parenchyma was segmented using a deep-learning model. MR images from 30 participants were registered using lung masks, modified masks or no weighting. Similarity metrics and functional parameters were compared. Results: Size-modified weighting masks significantly improved registration accuracy and ventilation agreement. Impact: The robustness of image registration is crucial for calculating ventilation in functional lung MRI. Selecting optimal weighting masks to support registration can significantly enhance precision, impacting the accuracy of functional ventilation post-processing.