Program Information
Validation of Time-Resolved 4DMRI by Using a Designed Ground Truth and Introducing the Concept of Heritage Consistency Error
G Li*, J Moody , A Sun , Memorial Sloan Kettering Cancer Center, New York, NY
Presentations
SU-K-605-2 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 605
Purpose: To validate the super-resolution approach to reconstruct time-resolved (TR) 4DMRI with a high spatiotemporal resolution by combining two inferior, yet complementary, image sets, using deformable image registration (DIR).
Methods: We have evaluated the image fidelity of TR-4DMRI by introducing a designed ground truth (GT) along with the concept of heritage consistency error (HCE). Six volunteers were scanned with the fast 3D acquisition protocol under an IRB-approved protocol. To assess DIR uncertainty and consistency, three sets of thorax-abdomen images were acquired for each of the six subjects: Two breath-holds (BHs) at full exhalation and full inhalation with high resolution (2x2x2mm³), and a series of free breathing (FB) low resolution (5x5x5mm³) 3D cine, at 2Hz. A Demons-based DIR algorithm was used to reconstruct TR-4DMRI. For validation, we generated a low-resolution BH1ᴸᵒʷᴿ from the high-resolution BH1, mimicking 3D cine, performed DIR of BH2-to-BH1ᴸᵒʷᴿ, and compared resulting TR-4DMRI image with the GT (original BH1), and vice versa. Corresponding anatomic points, such as diaphragm dome points and liver vessel bifurcation points, were used to assess DIR accuracy. We also compared the two TR-4DMRI images reconstructed from BH1-to-3Dcine and BH2-to-3Dcine, using HCE, which is defined as the voxel intensity difference (VID) of two resulting TR-4DMRI, normalized to the mean of the VIDs between each TR-4DMRI and corresponding 3D cine image.
Results: Good registration is achieved within 5mins for all six simulations. The accuracy for TR-4DMRI is 0.2±1.8mm in a high contrast region (the diaphragm) and ≤3±3mm in a low contrast region (the abdomen). The HCE value of DIR-based TR-4DMRI is 0.97±0.20, ranging from 0.8 to 1.3.
Conclusion: Our GT and HCE validation methods are useful in assessing TR-4DMRI image veracity and uncertainty. Additionally, the higher soft-tissue contrast in MRI results in higher DIR accuracy. Further improvement of TR-4DMRI image fidelity is currently under investigation.
Contact Email: