Room: Exhibit Hall | Forum 2
Purpose: To develop physiologically-plausible digital thoracic phantoms through respiratory motion modelling based on a combination of Hyperpolarized (HP) Gas tagging Magnetic Resonance Imaging (MRI) technique and Deformable Image Registration (DIR) of high-resolution proton MRI (2.5mm x 2.5mm x 2.5mm).
Methods: Lung images for three healthy subjects were acquired at end-of-inhalation (EOI) and end-of-exhalation (EOE) phases using a HP gas tagging MRI and a high-resolution proton MRI technique. The center of mass of the tagged grids was tracked from one phase to the other to obtain the ground-truth displacement vector field (DVF) within the lungs. The resolution of the physiologically-based sparse DVF, of approximately 270 grids, was increased through a cubic interpolation algorithm. A ventilation map was generated by calculating the Jacobian of this ground-truth DVF as a metric of lung function. In order to have a respiratory motion model of the entire thorax, the tagging-based DVF was used to optimize a DVF of the entire thorax obtained through DIR of the proton MRI. The optimization was performed by implementing a cost function that minimized the differences between the DVF from the tagging-MRI and the DVF from the proton-MRI, while preserving the ground-truth ventilation as the lung function metric in this study. To complete the digital phantom, a smoothing function which utilizes a Gaussian filter was applied to the combined DVFs in order to ensure continuity in the lung border regions.
Results: Physiologically-based thoracic digital phantoms were created based on optimization of DIR-based DVFs with ground-truth DVFs from a HP gas tagging MRI technique. The respiratory motion outside the lungs was modeled through DIR, and DVF continuity in the lung borders was successfully addressed through a smoothing algorithm.
Conclusion: The created digital phantoms provide a realistic respiratory motion model that can potentially be used as a validation tool for DIR standardization.
Funding Support, Disclosures, and Conflict of Interest: NIH grant: R21CA195317