Background
Quantitative myocardial T1 mapping is commonly performed using a breath-hold ECG-triggered acquisition. Despite breath-hold instructions, motion is observed in ~50% of patients due to diaphragmatic drift and their limited breath-holding capability [1]. Registration of each T1-weighted (T1w) image can be performed to reduce motion artifacts in T1 maps but remains challenging due to the high intensity variations among T1w images [1]. In this study we propose a novel non-rigid T1w image registration approach.
Methods
Our proposed method uses an extended formulation of the optical flow problem, where both motion field and intensity variation are estimated simultaneously within a unified variational framework [2]. An additional term was introduced to constrain the deformation field using automatic feature point tracking [3]. Each T1w image is registered to the 4th image of the series (reference), on which a region of interest is manually drawn around the left ventricle (LV-ROI). All remaining steps are performed automatically, where affine motion parameters are first estimated by maximization of the mutual information between the reference image and each T1w image over the LV-ROI, and is followed by our proposed non rigid motion estimation step. Twenty patients (57 ± 14 y, 12 m) referred for clinical CMR exams were scanned before and after administration of contrast agent. T1 mapping was performed in 1-3 slices with a 5-(3)-3 scheme for pre-contrast and 4-(1)-3-(1)-2 scheme for two post-contrast scans at ~15 and ~30 min post-injection. 85 total T1 maps were acquired and were visually assessed for the presence of motion. To quantify the registration step, the myocardium was manually segmented in all T1w images and the DICE coefficients were computed between each registered T1w image and the reference image (1: ideal match, 0: none). Overall T1 map quality and motion artifacts were assessed by a blinded reader using a 4-point scale (0: non diagnostic/severe motion artifact, 4: excellent/no motion artifact).
Results
57% of the T1w image series were visually identified as "with motion". After motion correction, DICE coefficients (Figure 1) were slightly improved in "no motion" series (0.90 ± 0.02 vs. 0.91 ± 0.02, p < 0.002) and greatly improved in "with motion" series (0.80 ± 0.14 vs. 0.89 ± 0.03, p < 0.001). Figure 2 shows T1 maps reconstructed with and without motion correction. No statistical difference was found in term of overall T1 map quality before and after correction in "no motion" series After motion correction, improved overall T1 map quality (2.86 ± 1.04 to 3.49 ± 0.77, p < 0.001) and reduced motion artifacts (2.51 ± 0.84 to 3.61 ± 0.64, p < 0.001) were obtained in "with motion" series.
Conclusions
The proposed non-rigid registration approach reduces the respiratory-induced motion occurring during breath-hold T1 mapping and significantly improves the quality of T1 maps.
Funding
NIH R01EB008743-01A2.
References
Xue: MRM. 2012
Cornelius: ACM-Siggraph. 1983
Roujol: IEEE-TITB. 2012
Author information
Authors and Affiliations
Rights and permissions
This article is published under an open access license. Please check the 'Copyright Information' section either on this page or in the PDF for details of this license and what re-use is permitted. If your intended use exceeds what is permitted by the license or if you are unable to locate the licence and re-use information, please contact the Rights and Permissions team.
About this article
Cite this article
Roujol, S., Foppa, M., Kawaji, K. et al. Improved motion correction for T1 mapping. J Cardiovasc Magn Reson 16 (Suppl 1), P45 (2014). https://doi.org/10.1186/1532-429X-16-S1-P45
Published:
DOI: https://doi.org/10.1186/1532-429X-16-S1-P45