Repeatability of Ventilation In the Lung Using MR-Based Imaging
Abstract
Purpose
In investigating image-based lung ventilation, a fundamental question must be answered: is the measurement technique repeatable? Repeatability of ventilation imaging in PET and CT have been studied, but MRI is emerging. Crucial to deriving ventilation is the deformable imaging registration (DIR) algorithm, and the algorithm choice can greatly influence the final ventilation. This work evaluates the repeatability of an MR-based ventilation imaging technique across three DIR algorithms to determine the most appropriate algorithm for MR-based ventilation imaging.
Methods
To investigate repeatability, three participants were each scanned twice, approximately an hour apart to allow a short break outside of the scanner, with a 1.5T MRI Siemens Magnetom Sola via a non-contrast protocol of the whole lung. Prospective gating using an ultra-short echo time stack-of-spirals sequence acquired k-space at distinct inhale and exhale phases, thus allowing participants to breathe freely. Three open-source DIR algorithms were used to measure exhalation to inhalation deformation vector fields (DVFs): dense displacement sampling (deeds), elastix, and MATLAB’s demons. The ventilation was found as the determinate of the Jacobian matrix of the DVF. Ventilation maps of intra-participant were compared across algorithms using the structural similarity index measure (SSIM), and the respective DIR algorithm was used to register exhalation of repeat scans to ensure consistent geometry for comparison.
Results
From repeated measurement pairs, elastix produced ventilation images with the highest SSIM. Averaged across participants, the SSIM of elastix-produced ventilation maps was 0.75±0.20. Qualitatively, elastix produced visually distinct ventilation between participants, furthering confidence in the technique.
Conclusion
Of the DIR algorithms investigated, elastix is most suitable for ventilation imaging with the non-contrast, whole lung, free-breathing MRI protocol presented here, demonstrating moderate repeatability in ventilation while maintaining clear visual differences between participants. Further work might expand the cohort and quantify inter-participant ventilation across DIR algorithms, as well as establish reproducibility.