Purpose: Utilizing a hands-free, MR-compatible 3D ultrasound transducer, we developed a method to bin multiphase 2D MR slices into MR volumes in time—effectively reconstructing a 4D MR dataset. The resulting dataset provides spatial and temporal information for image-guided radiotherapy applications.
Methods: Simultaneous multiphase 2D sagittal MR slices and 3D ultrasound images of the liver were acquired from four healthy volunteers as well as end-exhale 3D MR images on a clinical 3T MRI. The imaging session was repeated for two volunteers. Liver vessels in ultrasound were tracked throughout the session. Ultrasound motion datapoints were temporally matched to 2D MR images. Deviations from the mean greater than 1.5 standard deviations were rejected as outliers. The remaining datapoints were sorted into 14 equidistant amplitude bins. Likewise, datapoints were sorted according to MR slice position. Thus, subsets of images were created for each amplitude-slice position bin pair. The image with the median position within each subset was selected to represent the group. Selected images were reorganized to create static MR volumes at each amplitude. This technique and resulting reconstructions were evaluated quantitatively and qualitatively. End-exhale 3D MR provided ground-truth for comparison to the binned 3D MR at end-exhale as measured in diaphragm position discrepancy. Difference maps were assessed for volunteers with two imaging sessions.
Results: The average diaphragm position discrepancy was 2.14±0.92 mm. Qualitative analysis showed reasonable out-of-plane diaphragm continuity. Difference maps demonstrated good agreement, especially for the liver. Ultrasound tracking revealed extreme baseline drift in one dataset. The change in breathing motion was effectively captured with this system indicating this workflow can determine whether an imaging session recorded consistent, representative respiratory signals.
Conclusion: Simultaneous ultrasound was effectively used to reconstruct multiphase 2D MR images into 4D datasets. This technique could be applied to clinical datasets to assess target motion and deformation from respiration.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by NIH grant R01CA190298 as well as the University of Wisconsin-Madison Graduate School Pandemic Affected Research Continuation Initiative (PARCI).