Encrypted login | home

Program Information

A Technique to Generate Ultrafast Volumetric Cine MRI (UVC-MRI) Through Direct K-Space Matching for On-Board Target Localization

no image available
W Harris

W Harris*, C Wang , F Yin , L Ren , Duke University Medical Center, Durham, NC


SU-K-605-1 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 605

Purpose: To develop a technique to estimate on-board ultra-fast volumetric cine MRI (UVC-MRI) using sparsely sampled k-space cine data, prior 4D-MRI and direct k-space matching for real-time 3D target verification.

Methods: 4D-MRI volumes acquired during patient simulation were used as prior images. Respiratory motion models were built by structure principal component analysis of the deformation fields derived from 4D-MRI. On-board UVC-MRI at any instant was represented as a deformation of prior MRI at end of expiration phase. The deformation field was solved using the motion model and data fidelity constraint by matching to the sparsely acquired data directly in k-space. The UVC-MRI is finally generated by deforming prior MRI based on the deformation field solved. UVC-MRI based on direct k-space matching was evaluated using XCAT (computerized patient model) simulation of lung cancer patients with prior 4D-MRI and ground-truth on-board UVC-MRI simulated with various breathing pattern changes from simulation to treatment. Different sparse acquisition strategies, including Cartesian k-space sampling, radial k-space sampling and skipping-phase-encoded k-space sampling were investigated, and the results were compared with UVC-MRI generated by matching to a region-of-interest around the tumor in the sagittal 2D cine slice in image space.

Results: The average Volume-Percent-Difference(VPD)/Center-of-Mass-Shift(COMS) was 11.72±2.72%/0.91±0.41mm across all XCAT scenarios and sparse sampling acquisitions when matching directly to k-space in the UVC-MRI estimation. For Cartesian-based and radial-based sparse sampling, matching directly to k-space substantially reduced the reconstruction time, while maintaining overall UVC-MRI accuracy compared to matching to image-space. For skipped-phase-encoded-based sparse sampling, matching directly to k-space substantially improved the estimation accuracy of the UVC-MRI compared to matching to image-space from VPD/COMS of 24.45±27.89%/2.34±3.24mm to 13.04±4.17%/1.06±0.58mm averaged across all XCAT scenarios.

Conclusion: Preliminary studies showed that it is feasible to generate UVC-MRI by matching directly to k-space data using sparsely sampled cine acquisition for real-time 3D target localization.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by the National Institutes of Health Grant No. R01-CA184173 and a research grant from Varian Medical Systems.

Contact Email: