Synopsis

Motion correction with a model-target (MoCoMo) has been used in DCE-MRI to overcome the problem of changes in image contrast, but the method applies in principle to any other quantitative MRI method. The aim of this study is to demonstrate this hypothesis by applying the algorithm to renal DCE, DTI, T1 and T2-mapping in human subjects. The results show that MoCoMo is effective in removing even major motion effects in all 4 modalities and does not affect data where no motion is present. We conclude that MoCoMo is a suitable candidate for universal motion correction across all functional MRI modalities.

Introduction

Methods

Motion Correction with a Model Target (MoCoMo)

MoCoMo’s basic principle is illustrated in Fig 1(a). Given a set of N acquired frames, measured with N sets of imaging parameters, a new set of N registered frames is created by iterating two steps: (1) the current set of N registered frames is fitted pixel-by-pixel to the signal model, producing N fitted frames; (2) these are used as targets for a frame-by-frame registration of the N acquired frames, producing a new set of N registered frames. In the first iteration the N registered frames are initialised to the N acquired frames.

Image acquisition

Test data were drawn from renal pilot studies for 4 different contrast mechanisms, measured in 3 centres, on 2 different 3T scanners (Siemens, Philips). Fig.1(b) presents the basic acquisition parameters. 3 datasets from different subjects were randomly selected for each of the 4 contrast mechanisms.

Image processing

Data were processed by the same observer using in-house built Matlab algorithms. 2D registrations of a single slice were performed for this study. Pixel-by-pixel model fitting was performed with a linear 2-compartment model fit for DCE-MRI⁽³⁾, a mono-exponential recovery for T1⁽⁴⁾, a mono-exponential decay for T2⁽⁵⁾, and a diffusion tensor model for DTI⁽⁶⁾. For the DCE, an arterial input function was selected manually inside the abdominal aorta on the transverse slice. Frame-by-frame image co-registration was implemented in Elastix⁽⁷⁾ following a multi-resolution scheme, using a mutual information metric with adaptive stochastic gradient descent⁽⁸⁾. Each registration was performed with a combination of rigid body registration (Euler, to correct for gross motion) and free-form registration (B-spline, for smaller deformations). The control points for B-spline were placed 100 mm apart for the first iteration, and reduced stepwise to 25mm over the subsequent iterations. 3 iterations were performed in total.

Evaluation of results

The effects of MoCoMo were evaluated by visual comparison of parameter maps calculated with and without motion correction. For each contrast mechanism, evaluation was based on the parameter most sensitive to motion: T1, T2, Fractional Anisotropy (FA) from DTI and Blood Flow (BF) from DCE.

Results

Fig.2-5 show the results without motion correction (top row) and with motion correction (bottom row) for each of the four contrast mechanisms (T1, T2, FA, BF respectively), for each subject (left, middle, right column). Visual inspection of the T1 and T2-mapping data acquired during breath hold showed image artefacts due to imperfect rephrasing in the readout, but little to no motion-induced displacements. The free-breathing DTI and DCE data showed heavy breathing motion, while the shallow breathing DCE data showed low-amplitude breathing motion.

The results show that:

1. in T1, T2-mapping the difference between corrected and uncorrected parameter maps is negligible;
2. in shallow breathing DCE data (Fig.5 columns2-3), MoCoMo produces sharpened parameter maps with increased BF values in the renal cortex, consistent with removal of motion-induced partial volume effects;
3. in heavy breathing DCE and DTI data, MoCoMo shows a major improvement on anatomical delineation and precision of parameter maps.

Discussion

Conclusion

Acknowledgements

References

1. Buonaccorsi GA, O'Connor JP, Caunce A, Roberts C, Cheung S, Watson Y, et al. Tracer kinetic model–driven registration for dynamic contrast‐enhanced MRI time‐series data. Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine. 2007;58(5):1010-9.

2. Adluru G, DiBella EV, Schabel MC. Model-based registration for dynamic cardiac perfusion MRI. J Magn Reson Imaging. 2006;24(5):1062-70.

3. Flouri D, Lesnic D, Sourbron SP. Fitting the two-compartment model in DCE-MRI by linear inversion. Magn Reson Med. 2016;76(3):998-1006.

4. Messroghli DR, Radjenovic A, Kozerke S, Higgins DM, Sivananthan MU, Ridgway JP. Modified Look-Locker inversion recovery (MOLLI) for high-resolution T1 mapping of the heart. Magn Reson Med. 2004;52(1):141-6.

5. De Bazelaire CM, Duhamel GD, Rofsky NM, Alsop DC. MR imaging relaxation times of abdominal and pelvic tissues measured in vivo at 3.0 T: preliminary results. Radiology. 2004;230(3):652-9.

6. Kingsley PB. Introduction to diffusion tensor imaging mathematics: Part II. Anisotropy, diffusion-weighting factors, and gradient encoding schemes. Concepts in Magnetic Resonance Part A. 2006;28A(2):123-54.

7. Klein S, Staring M, Murphy K, Viergever MA, Pluim JP. elastix: a toolbox for intensity-based medical image registration. IEEE Trans Med Imaging. 2010;29(1):196-205.

8. Staring M, Klein S, Reiber JH, Niessen WJ, Stoel BC. Pulmonary image registration with elastix using a standard intensity-based algorithm. Medical Image Analysis for the Clinic: A Grand Challenge. 2010:73-9.