Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Nov 1.
Published in final edited form as: Magn Reson Med. 2020 Apr 17;84(5):2400–2411. doi: 10.1002/mrm.28283

Three Dimensional Motion Corrected T1 Relaxometry with MPnRAGE

Steven Kecskemeti 1,2,*, Andrew L Alexander 1,3,4
PMCID: PMC7396302  NIHMSID: NIHMS1587567  PMID: 32301173

Abstract

Purpose:

To test the performance of the MPnRAGE motion correction algorithm on quantitative relaxometry estimates.

Methods:

Twelve children (9.4+/−2.6 years, min=6.5 years, max = 13.8 years) were imaged three times in a session without sedation. Stabilization padding was not used for the second and third scans. Quantitative T1 values were estimated in each voxel on images reconstructed with and without motion correction. Mean T1 values were assessed in various regions determined from automated segmentation algorithms. Statistical tests were performed on mean values and the coefficient of variation across the measurements. Accuracy of T1 estimates were determined by scanning the High Precision Devices (HPD) MRI System phantom with the same protocol.

Results:

T1 values obtained with MPnRAGE agreed within 4% of the reference values of the HPD phantom. The best fit line was T1(MPnRAGE) = 1.02 T1(reference) – 0.9 ms, R2=0.9999. For in vivo studies, motion correction reduced the coefficients of variation of mean T1 values in whole brain tissue regions determined by FSL FAST by 74+/−7%, subcortical regions determined by FIRST and FreeSurfer by 32+/−21% and 33+/−26% respectively. Across all participants, the mean coefficients of variations ranged from 0.8 to 2.0% for subcortical regions and 0.6%+/−0.5% for cortical regions when motion correction was applied.

Conclusions:

MPnRAGE demonstrated highly accurate values in phantom measurements. When combined with retrospective motion correction, MPnRAGE demonstrated highly reproducible T1 values, even in participants who moved during the acquisition.

Keywords: Motion correction, Relaxometry, T1, R1, MPnRAGE

Introduction

Quantitative relaxometry imaging methods have been used to investigate changes in brain tissues ranging from development1–7, aging, myelination and pathology. The general approach common to all methods is the collection of a series of images with different acquisition parameters, i.e. flip angles, repetition times (TRs), inversion times (TIs), etc. in order to estimate the underlying parameters. This may also include additional “calibration” scans to measure scanner imperfections such as scans to map the magnetic field inhomogeneity (B0 map) or radio-frequency transmit field (B1map). Head motion, both during individual scans or between the different scans poses a challenge to robust parameter estimation. To date, almost all of the effort in reducing the effects of motion in quantitative relaxometry have been in registration of the individual source images. This works well when the individual images are not corrupted by motion artifacts and the relative motion between scans is minimal. However, spatial interpolations used in this step can produce unwanted smoothing effects and inaccuracies 8. Larger motions between the individual scans may be problematic for high density receiver arrays with rapidly varying spatial sensitivities since the signal at each anatomical location is then modulated with a different coil sensitivity in each of the images. Since the receiver sensitivity is typically incorporated into an effective “spin density”, this effectively creates an additional unknown parameter with each new motion event. The gross effects can largely be reduced using intensity correction methods 9,10, but subtle effects are likely to persist. Motion during the scans results in corrupt images, such as the ghosting artifacts when Cartesian k-space sampling is used or image blurring when radial k-space sampling is used 11. One strategy for dealing with motion is to acquire redundant data, i.e. more than the minimum number of flip angles, TIs, etc. required to perform fitting in anticipation of rejecting motion corrupted images during a quality control step of the processing; however, this increases the minimum scan time and decreases the measurement efficiency.

Quantitative relaxometry, especially quantitative T1 in children is a highly desirable method to track normal and abnormal brain development as myelin content is believed to be a major factor contributing to T1 12. However, accurate and stable methods needed to track such changes are difficult to achieve in children as motion effects are more dominant than in adults. Although sedation may be used to reduce motion, it is generally not feasible in research studies and many clinical centers are trying to minimize the use for clinical studies due to potential side effects.13 In studies of very young children, imaging during natural sleep has been used as a strategy for minimizing motion 14,15; however, motion can still occur and the strategy becomes less effective for older children. Prospective motion correction using optical tracking, imaging navigators, and FID navigators are also promising strategies 16,17, but remain largely unvalidated at correcting motion in quantitative relaxometry18. Retrospective motion correction methods have recently been developed for low resolution 2D T1 and T2 relaxometry mapping with MR fingerprinting19. To date, there are few demonstrations of motion corrected relaxometry with 3D, high resolution, whole brain methods.18

Retrospective motion correction using self-navigation with k-space blades (i.e., PROPELLER –20) or 3D radial k-space sampling are also promising strategies for motion correction.21 Recently, an inversion-prepared 3D radial imaging sequence, MPnRAGE, was proposed as a novel technique for simultaneously generating multiple inversion-recovery contrasts and quantitative T1 imaging22. Subsequently, the reconstruction for this pulse sequence was adapted to enable retrospective motion correction using low resolution images reconstructed from the measurements following each inversion recovery pulse23. MPnRAGE motion correction has been demonstrated to generate consistently high quality whole-brain T1-weighted images in a neuroimaging study of children with autism and typical development23. More recently, high test-retest performance of automated segmentations was demonstrated on a cohort of unsedated pediatric children imaged with motion corrected MPnRAGE.24 Since retrospective motion correction with MPnRAGE was shown to consistently provide high quality T1-weighted images, we hypothesized that these methods would also enable improved quality and consistency of quantitative T1 relaxometry from the same data. In this study, retrospective motion-corrected T1 relaxometry with MPnRAGE was evaluated in a test-retest study of twelve children scanned without sedation.

Methods

Study Population

Imaging experiments were performed with institutional review board approval and informed consent/assent. Twelve children (9.4+/−2.6 years, min=6.5 years, max=13.8 years, 6 male and 6 female) without known neurological health concerns were selected for imaging. Recruitment was not based on likelihood of subjects remaining still during the scan.

Image Acquisition

All exams took place on a 3T MRI scanner (Discovery MR750, GE Healthcare, Waukesha, WI) without the use of sedation. Participants watched a video of their choice and were instructed to remain still. The participants heads were stabilized within a 32 channel head coil (Nova Medical, Wilmington, MA) using the NoMoCo pillow support system (NoMoCo Pillow, Inc., La Jolla, Ca). After receiving a single MPnRAGE acquisition and prospectively corrected T1-weighted MPRAGE acquisition (6-7 minutes), some of the padding was removed to allow further range of motion. Each subject then received two additional MPnRAGE and MPRAGE acquisitions in alternating order. The order of all scans was counter-balanced across subjects. The prospectively corrected MPRAGE acquisitions were not used in this analysis as they did not provide T1 estimates.

MPnRAGE Background

A magnetization (inversion) prepared rapid gradient echo (i.e. MPRAGE25) sequence was modified to collect a large number, n, gradient echoes (i.e. MPnRAGE23) with a 3D radial k-space trajectory after each preparation pulse. The trajectory ordering was quasi-randomly ordered to approximately distribute projection angles uniformly within each gradient echo block as well as across each gradient echo block, at each inversion time. Low resolution navigator images are formed using data within each gradient echo block to estimate motion parameters. Motion correction is performed directly in k-space using the Fourier Shift Theorem to correct translations and by rotating the k-space trajectories to correct for rotations. Thus, errors from spatial interpolations are eliminated. Differently contrasted images (i.e. inversion time images) are formed using the collection of data at a specific inversion time, but across all gradient echo blocks and are used to estimate T1 by fitting to a known model, derived by solving the Bloch equations.

MPnRAGE Acquisition Parameters:

Whole brain coverage with 1.0 mm isotropic resolution was acquired in the axial orientation with 200 slices. Parameters included, TR=4.9 ms, TE=1.8 ms, n=386 views along the recovery curve. The excitation flip angles were 4°/8° for the first 325/remaining 61 views. A delay time of TD=500 ms occurred after the last TR of each gradient echo block to allow the signal to recover before the next preparation pulse. The scan time was 7 minutes.

MPnRAGE Reconstruction and T1 Fitting Procedure:

Composite T1-weighted images using the entire k-space dataset for each scan were reconstructed both with and without motion correction. Complex valued source images corresponding to each of the 386 inversion times (12ms to 1889ms, evenly spaced by 1TR) were reconstructed with and without the motion correction algorithm described in23 by summing the individual coil images according to 26. A two pass fitting procedure similar to 27 was performed that initially fit all voxels for 4 unknowns (T1, spin density, B1, and inversion efficiency) using a known model determined by solving the Bloch equations. The B1 and inversion efficiency maps were smoothed using a 3D-Gaussian kernel with full-width half-max at 7 and 9 mm, respectively. The T1 and spin density maps were then refit while fixing the model using the smoothed B1 and inversion efficiency maps. The T1 maps from the second fitting pass were then denoised using total variation (TV) minimization.

Phantom Scans:

The High Precision Devices (HPD) system phantom (Model 130) was used to assess the accuracy of estimated T1 values from MPnRAGE. The HPD system phantom developed by a joint effort between the National Institute of Standards and Technology (NIST) and the International Society for Magnetic Resonance in Medicine (ISMRM), contains 14 spherical samples with a range of known T1 values 28. All image acquisition, reconstruction, and fitting procedures were the same as the in-vivo experiments. Mean T1-values of each phantom were compared to the known values using linear regression.

ROI selection & T1 measurement:

The T1-weighed composite images for each subject were used as inputs to fsl_anat from the FMRIB Software Library v5.0 and recon-all from FreeSurfer. “fsl_anat” is a wrapper provided in the FMRIB Software Library that performs biasfield correction, brain and lesion extraction, whole brain white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) segmentations with FAST 29 (a hidden Markov random field model), and subcortical segmentations with FIRST 30 (a Bayesian approach of shape and appearance). In addition, the recon-all tool from FreeSurfer generated a cortical surface based segmentation 31 and subcortical segmentation with a Bayesian approach using a realistic image likelihood term as well as a prior model 32. The default settings were used for both programs, although the images were initially corrected for the strong receiver biases using N4BiasFieldCorrection, part of the ANTS software package and described in 9, before calling recon-all. For analysis, the whole brain white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF) masks from the FAST segmentations of fsl_anat, as well as the subcortical masks for the thalamus-proper, caudate, putamen, pallidum, hippocampus, amygdala, accumbens-area from both hemispheres and the brain-stem/4th ventricle were obtained from FIRST and FreeSurfer automated segmentations.

Since the quantitative T1 maps are inherently aligned with the T1-weighted composite images, no additional registration steps are needed for alignment with the segmentation masks other than to correct for some 90 degree rotations introduced by each method needed for rough alignment with their prior images, which may be corrected for with FSL’s fslreorient2std and FreeSurfer’s mri_convert utilities. The T1 values of all voxels within each region for each scan and subject are saved and used for analysis. Values within the cortical surface determined by FreeSurfer are sampled at the midpoint between the white matter / gray matter boundary and the pial surface using FreeSurfer’s tool mri_vol2surf with option “--projfrac 0.5”.

Prior Analyses

The T1-weighted composite images were previously analyzed in 24 , which demonstrated high test-retest of the automated segmentation algorithms used throughout this study.

Analyses

For region of interest (ROI) based analyses, the stability of the mean T1 value within each ROI across repeated measurements within a subject is assessed using the coefficient of variation (COV). The mean and standard deviations of the COV across all subjects was computed for each ROI. Statistical differences in the mean and standard deviation of the COVs with and without correction were assessed using paired-sample t-tests and two-sampled F-tests. a Bonferroni multiple comparison correction was used to adjust the significance level to p=0.05/m, where m is the number of measurements for each method.

Results

The results from the HPC phantom scan are shown in Table 1. Across a wide range of values, encompassing those expected in the human brain (T1 values between about 2000ms and 350ms), MPnRAGE differed from the reference values less than 4%. The best fit line was T1(MPnRAGE) = 1.02 T1(reference) – 0.9 ms, R2=0.9999 using phantoms T1-1 through T1-13. T1 estimation for phantom T1-14, with T1~28ms, was off by about 7 ms (23%).

Table 1:

Estimated T1 values from the HPC System Phantom. Excluding phantom T1-14, the best fit line was T1(MPnRAGE) = 1.02 T1(reference) – 0.9 ms, R2=0.9999

Sample Name Reference (ms) MPnRAGE (ms) MPnRAGE (% error)
T1-1 1989 ± 1.0 2021 ± 270 1.6 %
T1-2 1454 ± 2.5 1497 ± 130 3.0 %
T1-3 984.1 ± 0.33 1005 ± 54 2.1%
T1-4 706 ± 1.5 730 ± 33 3.4%
T1-5 496.7 ± 0.41 506 ± 28 1.9%
T1-6 351.5 ± 0.91 351 ± 21 −0.3%
T1-7 247.13 ± 0.086 256 ± 12 3.4%
T1-8 175.3 ± 0.11 172 ± 7 −2.21%
T1-9 125.9 ± 0.33 129 ± 6 2.3 %
T1-10 89.0 ± 0.17 92 ± 5 2.8 %
T1-11 62.7 ± 0.13 63 ± 6 0.8 %
T1-12 44.53 ± 0.090 42 ± 6 −5.1 %
T1-13 30.84 ± 0.016 30 ± 7 −1.7 %
T1-14 27.719 ± 0.0054 21 ± 6 −23%
Open in a new tab

Example T1 weighted images and T1 maps and estimated motion parameters are shown in Figures 1–3 for subjects demonstrating negligible/minor (Fig 1), moderate (Fig 2), and severe (Fig 3) motion artifacts. Supporting Information Videos S1 and S2 contain six examples in animated GIF format illustrating the differences in T1-weighted and quantitative T1 maps with and without motion correction for cases with mild, moderate, and severe motion artifacts. Figure 1 demonstrates the ability to improve even minor motion artifacts that are hardly noticeable on the uncorrected images. Moderate motions in Figure 2 are nearly eliminated after motion correction, while artifacts from severe motions in Fig. 3 are considerably reduced, but not necessarily entirely eliminated.

Figure 1:

Figure 1:

Open in a new tab

T1-weighted images and quantitative T1 maps from a 79 month old boy demonstrating ability of MPnRAGE to correct for subtle motion artifacts.

Figure 3:

Figure 3:

Open in a new tab

T1-weighted images and quantitative T1 maps from a 92 month old female demonstrating ability of MPnRAGE to correct for severe motion artifacts arising from nearly continuous motion events throughout the duration of the scan.

Figure 2:

Figure 2:

Open in a new tab

T1-weighted images and quantitative T1 maps from a 108 month old boy demonstrating ability of MPnRAGE to correct for moderate motion artifacts.

The results of T1 reliability for whole brain WM, GM, and CSF segmentations from FAST are shown in Figure 4 and Supporting Information Table S1. All three tissue types had statistically significant differences in the mean between the motion corrected and non motion corrected T1 values, but only CSF had a statistically significant difference of the standard deviation. With motion correction, the mean and standard deviations of the COVs for all regions were significantly reduced (P<0.05/3), with motion correction offering a profound decrease in the COV (1.3% vs 7.5% for CSF, 0.5% vs 1.9% for GM and 0.7% vs 2.4% WM). Compared to the values without motion correction, the COVs with motion correction are on average 74%/−7% less. Bland-Altman plots of the T1 value differences with and without corrections are presented in Figure 5 (color coded by region of interest) and Supporting Information Figure S1 (color coded by subject) for all measurements of all subjects. Most of the measurements with and without correction were relatively consistent although larger differences were observed across multiple regions and cases.

Figure 4:

Figure 4:

Open in a new tab

T1 values (top) and coefficients of variation (bottom) from regional measurements in whole brain white cerebrospinal fluid, gray matter, and white matter segmentations from FSL FAST across all subjects. Superscripts t,f denote p<0.05 for t-tests and f-tests, used to test for statistical differences of the means (t-tests) and standard deviations (f-tests) between the motion corrected and non-motion corrected T1 values. The T1 values for CSF have been multiplied by 0.5 to preserve dynamic range of the plots.

Figure 5:

Figure 5:

Open in a new tab

A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FAST. Shown are all measurements from all subjects and trials, color coded by region of interest. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

The results from the FSL FIRST and FreeSurfer segmentations are presented in Figures 6–8 and Supporting Information Tables S2–S5. Seven of fifteen regions in FIRST and seven of fifteen regions in FreeSurfer had significantly different T1 values when measured with and without motion correction after corrections for multiple comparisons (P<0.05/15). With motion correction, the average COV was reduced for all regions except the right thalamus. The mean values of the regional COVs after motion correction were between 0.8–2.0%. Compared to the values without motion correction, the COVs after motion correction are lower on average 32%+/−21% and 33%+/−25% for FSL FIRST and FreeSurfer, respectively. The reduction in COV was judged statistically significant for at least one segmentation method for all regions except the left and right amygdala and palladium. However, only the left hippocampus were significant at the multiple-comparisons correction level of P<0.05/15. Bland-Altman plots of the T1 values with and without are presented in Figure 8 (color coded by region of interest) and Supporting Information Figures S2–S3 (color coded by subject) for all measurements of all subjects. The mean T1 values increased after motion correction by about 1% for each method, although both methods had about one third (32% and 35% or FIRST and FreeSurfer) of the cases report decreases.

Figure 6:

Figure 6:

Open in a new tab

T1 values (top) and coefficients of variation (bottom) from regional measurements in subcortical segmentations from FSL FIRST across all subjects. Superscripts t,f denote p<0.05 for t-tests and f-tests, used to test for statistical differences of the means (t-tests) and standard deviations (f-tests) between the motion corrected and non-motion corrected T1 values.

Figure 8:

Figure 8:

Open in a new tab

A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FIRST and FREESURFER. Shown are all measurements from all subjects and trials, color coded by region of interest. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

Mean cortical T1 values are increased from 1539 ms ± 80 ms without motion correction to 1615 ms ± 40 ms with motion correction. A paired sample t-test confirms that this shift is significant (p=0.004) and a f-test confirms that the standard deviations are statistically different (p=0.03). When averaged across all subjects, the COV of mean cortical T1 was 4% ± 4% without motion correction and 0.6% ± 0.5% with motion correction.

Discussion

This study demonstrated that MPnRAGE yields highly accurate and reliable quantitative T1 mapping. The quantitative T1 mapping with MPnRAGE was found to be remarkedly consistent with the gold standard NIST calibrated T1 phantom 28. Similar results were observed with in vivo comparisons between MPnRAGE T1 mapping and a 2D inversion recovery quantitative T1 mapping method 27. Even when child participants were not sedated and did not have their head well stabilized within the head coil, MPnRAGE demonstrated high reproducible T1 values within automatically labeled regions. When motion correction was applied, not only was the COV reduced in almost all regions, but the T1 values also showed small, albeit statistical differences illustrating the importance of motion correction.

Retrospective motion correction was found to be highly effective for improving the reliability of quantitative T1 mapping in young children. We note that motion-corrected MPnRAGE was 100% successful in providing images for automated brain segmentation without major failure. While the study did not investigate specific types or scripted motion, the method appeared to be highly robust to a broad range of motion amplitudes and types (sharp transitions, jittery, and slow variations). This study was performed in children as young as 6 years of age without head restraints so it is anticipated that the reliability will be even better for more compliant cohorts with padding to help minimize head motion. A limitation is that the small sample size is not sufficient to fully characterize the performance as a function of the motion behavior.

The low COV of T1 with motion-corrected MPnRAGE, even in this challenging cohort, indicates that MPnRAGE will be an outstanding tool for quantitative brain imaging studies. Lower variability in T1 estimation will improve statistical power of statistical detection of small brain changes, particularly for within-subject longitudinal studies. A prior work 18 used a prospective motion correction approach with a motion camera and Moire phase marker to demonstrate reduced coefficients of variation in 3D, high resolution quantitative R1 relaxometry and magnetization transfer (MT). That approach offered a considerable reduction in median COVs, however, the resulting median COV for R1 with motion correction was still high (~>15%) and had pronounced artifacts in the motion corrected R1 maps. Since the no motion case had median COV about 13%, it is not clear if the large COV was due to motion, or perhaps an aggressive protocol consisting of 0.8 mm isotropic resolution that would suffer from inherently lower SNR than a 1.0 mm protocol. A separate study used the same relaxometry method without motion correction in a cohort of adult control subjects and found the COVs to be about 4 to 5 percent when 1.0 mm isotropic spatial resolution was used 33. In another study of quantitative T1 relaxometry of adult control subjects (23.2+/−3.6yrs) without motion correction and with 1.2 mm isotropic resolution, the mean intrasubject COVs was reported at less than 1% 34. In contrast, this manuscript reported COVs of 1–2% for 1mm T1 relaxometry from pediatric participants scanned without stabilization pads.

Although statistical differences between T1 maps with and without motion correction exist in most regions, one somewhat surprising finding in this work was the relative similarities of the quantitative T1 values. This is likely due to the inherent insensitivities of radial k-space trajectories to motion, which tend to produce a slight blurring artifact as opposed to the traditional ghost artifact with Cartesian k-space trajectories 11. Since the T1 and proton density values within a tissue region are expected to be highly homogenous, and since other parameters like B1 and inversion efficiency are expected to slowly vary across space, the parameter fitting is likely most affected near regional boundaries. Thus, voxels near regional boundaries may produce inaccurate parameter estimates when motion is not corrected. Reduced resolution may also hinder automated segmentation methods, causing inaccurate segmentations. We should also note that although FIRST and FreeSurfer are widely used on pediatric populations there has not be extensive validation or comparison with manual segmentations. One recent manuscript reported lower agreement between automatically segmented regions from FIRST and FreeSurfer to manually segmented regions of hippocampus and amygdala volumes from pediatric participants, although the authors could not exclude motion artifacts from contributing to the quality of the automated segmentations35. Without performing manual segmentations, it is difficult to determine if the observed statistical changes in T1 without motion correction are due to tissue signal mixing, inaccurate boundaries/segmentations, or a combination of the two.

Conclusions

MPnRAGE demonstrated highly accurate quantitative T1 relaxometry as assessed with HPD System phantom. A test-retest study of 12 pediatric participants scanned without sedation and stabilization pads was used to demonstrate highly reproducible T1 values, as assessed with region of interest measurements determined from automated segmentation techniques. Statistical differences were observed between T1 values obtained with and without motion correction, suggesting the importance of motion correction when performing quantitative relaxometry.

Supplementary Material

Supp Video 1

Supporting Information Video S1: This animated GIF shows six examples of T1-weighted images from MPnRAGE reconstructed with and without motion correction. The top row shows two cases with mild motion artifacts, the middle row shows two cases with moderate motion artifacts, and the bottom row shows two cases with severe motion artifacts. In all cases retrospective motion correction improves visual sharpness of images.

Supp Video 2

Supporting Information Video S2: This animated GIF shows six examples of quantitative T1-relaxometry images from MPnRAGE reconstructed with and without motion correction. The top row shows two cases with mild motion artifacts, the middle row shows two cases with moderate motion artifacts, and the bottom row shows two cases with severe motion artifacts. In all cases retrospective motion correction improves visual sharpness of images. These are the same cases and slices from Supporting Information Video S1.

Supplement

Supporting Information Table S1: Mean T1 values and coefficients of variation (COV) for region of interest measurements from FSL FAST. Compared to the values without motion correction, the COVs with motion correction are on average 74% +/− 7% less. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/3 = 0.0167 are highlighted in gray.

Supporting Information Table S2: The coefficient of variation (COV) for region of interest measurements of T1 from FSL FIRST. Compared to the values without motion correction, the COVs after motion correction are on average 32% +/− 21% less. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Table S3: The coefficient of variation (COV) for region of interest measurements of T1 from FreeSurfer. Compared to the values without motion correction, the COVs with motion correction are on average 33% +/− 26% less. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Table S4: The mean T1 values for region of interest measurements from FSL FIRST. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Table S5: The mean T1 values for region of interest measurements from FreeSurfer. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Figure S1: A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FAST. Shown are all measurements from all subjects and trials, color coded by subject. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

Supporting Information Figure S2: A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FIRST. Shown are all measurements from all subjects and trials, color coded by subject. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

Supporting Information Figure S3: A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FreeSurfer. Shown are all measurements from all subjects and trials, color coded by subject. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

Figure 7:

Figure 7:

Open in a new tab

T1 values (top) and coefficients of variation (bottom) from regional measurements in subcortical segmentations from FreeSurfer across all subjects. Superscripts t,f denote p<0.05 for t-tests and f-tests, used to test for statistical differences of the means (t-tests) and standard deviations (f-tests) between the motion corrected and non-motion corrected T1 values.

References

  • 1.Dean DC, Hurley SA, Kecskemeti SR, O’Grady JP, Canada C, Davenport-Sis NJ, et al. Association of amyloid pathology with myelin alteration in preclinical Alzheimer disease. JAMA Neurology. 2017;74(1). doi: 10.1001/jamaneurol.2016.3232 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Dean DC, Planalp EM, Wooten W, Adluru N, Kecskemeti SR, Frye C, et al. Mapping White Matter Microstructure in the One Month Human Brain. Scientific Reports. 2017;7(1). doi: 10.1038/s41598-017-09915-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Dean DC, Planalp EM, Wooten W, Schmidt CK, Kecskemeti SR, Frye C, et al. Investigation of brain structure in the 1-month infant. Brain Structure and Function. 2018. doi: 10.1007/s00429-017-1600-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Dean DC, Sojkova J, Hurley S, Kecskemeti S, Okonkwo O, Bendlin B, et al. Alterations of myelin content in Parkinson’s disease:a cross-sectional neuroimaging study. PLoS ONE. 2016;11(10). doi: 10.1371/journal.pone.0163774 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Leppert IR, Almli CR, McKinstry RC, Mulkern RV, Pierpaoli C, Rivkin MJ, et al. T(2) relaxometry of normal pediatric brain development. Journal of magnetic resonance imaging : JMRI. 2009;29(2):258–267. doi: 10.1002/jmri.21646 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Cheng H-LM, Stikov N, Ghugre NR, Wright GA. Practical medical applications of quantitative MR relaxometry. Journal of magnetic resonance imaging : JMRI. 2012;36(4):805–824. doi: 10.1002/jmri.23718 [DOI] [PubMed] [Google Scholar]
  • 7.Watanabe M, Liao JH, Jara H, Sakai O. Multispectral quantitative MR imaging of the human brain: lifetime age-related effects. Radiographics : a review publication of the Radiological Society of North America, Inc 2013;33(5):1305–1319. doi: 10.1148/rg.335125212 [DOI] [PubMed] [Google Scholar]
  • 8.Nachmani A, Schurr R, Joskowicz L, Mezer AA. The effect of motion correction interpolation on quantitative T1 mapping with MRI. Medical image analysis. 2019;52:119–127. doi: 10.1016/j.media.2018.11.012 [DOI] [PubMed] [Google Scholar]
  • 9.Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, et al. N4ITK: improved N3 bias correction. IEEE transactions on medical imaging. 2010;29(6):1310–1320. doi: 10.1109/TMI.2010.2046908 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Lorio S, Tierney TM, McDowell A, Arthurs OJ, Lutti A, Weiskopf N, et al. Flexible proton density (PD) mapping using multi-contrast variable flip angle (VFA) data. NeuroImage. 2019;186:464–475. doi: 10.1016/j.neuroimage.2018.11.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Glover GH, Pauly JM. Projection Reconstruction Techniques for Reduction of Motion Effects in MRI. Magnetic Resonance in Medicine. 1992;28(2):275–289. doi: 10.1002/mrm.1910280209 [DOI] [PubMed] [Google Scholar]
  • 12.Heath F, Hurley SA, Johansen-Berg H, Sampaio-Baptista C. Advances in noninvasive myelin imaging. Developmental neurobiology. 2018;78(2):136–151. doi: 10.1002/dneu.22552 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Malviya S, Voepel-Lewis T, Prochaska G, Tait AR. Prolonged recovery and delayed side effects of sedation for diagnostic imaging studies in children. Pediatrics. 2000;105(3):E42. doi: 10.1542/peds.105.3.e42 [DOI] [PubMed] [Google Scholar]
  • 14.Dean DC 3rd, Dirks H, O’Muircheartaigh J, Walker L, Jerskey BA, Lehman K, et al. Pediatric neuroimaging using magnetic resonance imaging during non-sedated sleep. Pediatric radiology. 2014;44(1):64–72. doi: 10.1007/s00247-013-2752-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Rosenberg DR, Sweeney JA, Gillen JS, Kim J, Varanelli MJ, O’Hearn KM, et al. Magnetic resonance imaging of children without sedation: preparation with simulation. Journal of the American Academy of Child and Adolescent Psychiatry. 1997;36(6):853–859. doi: 10.1097/00004583-199706000-00024 [DOI] [PubMed] [Google Scholar]
  • 16.Forman C, Aksoy M, Hornegger J, Bammer R. Self-encoded marker for optical prospective head motion correction in MRI. Medical image analysis. 2011;15(5):708–719. doi: 10.1016/j.media.2011.05.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Maclaren J, Armstrong BSR, Barrows RT, Danishad KA, Ernst T, Foster CL, et al. Measurement and correction of microscopic head motion during magnetic resonance imaging of the brain. PloS one. 2012;7(11):e48088. doi: 10.1371/journal.pone.0048088 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Callaghan MF, Josephs O, Herbst M, Zaitsev M, Todd N, Weiskopf N. An evaluation of prospective motion correction (PMC) for high resolution quantitative MRI. Frontiers in Neuroscience. 2015;9:97. doi: 10.3389/fnins.2015.00097 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Cruz G, Jaubert O, Schneider T, Botnar RM, Prieto C. Rigid motion-corrected magnetic resonance fingerprinting. Magnetic resonance in medicine. 2019;81(2):947–961. doi: 10.1002/mrm.27448 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Pipe JP. Motion correction with PROPELLER MRI: Application to head motion and free-breathing cardiac imaging. Magnetic Resonance in Medicine. 1999;42(5):963–969. [DOI] [PubMed] [Google Scholar]
  • 21.Anderson AG, Velikina J, Block W, Wieben O, Samsonov A. Adaptive retrospective correction of motion artifacts in cranial MRI with multicoil three-dimensional radial acquisitions. Magnetic Resonance in Medicine. 2013;69(4):1094–1103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Kecskemeti S, Samsonov A, Hurley SA, Dean DC, Field A, Alexander AL. MPnRAGE: A technique to simultaneously acquire hundreds of differently contrasted MPRAGE images with applications to quantitative T1 mapping. Magnetic Resonance in Medicine. 2016;75(3):1040–1053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Kecskemeti S, Samsonov A, Velikina J, Field AS, Tursky P, Rowley H, et al. Robust Motion Correction Strategy for Structural MRI in Unsedated Children Demonstrated with Three-dimensional Radial MPnRAGE. Radiology. 2018;289(2):509–516. doi: 10.1148/radiol.2018180180 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Kecskemeti SR, Alexander AL. Test-retest of automated segmentation with different motion correction strategies: A comparison of prospective versus retrospective methods. Neuroimage. 2019;209:116494. doi: 10.1016/j.neuroimage.2019.116494 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Mugler JP, Brookeman JR. Three-dimensional magnetization-prepared rapid gradient-echo imaging (3D MP RAGE). Magnetic Resonance in Medicine. 1990;15(1):152–157. doi: 10.1002/mrm.1910150117 [DOI] [PubMed] [Google Scholar]
  • 26.Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased array. Magnetic Resonance in Medicine. 1990;16(2):192–225. [DOI] [PubMed] [Google Scholar]
  • 27.Kecskemeti S, Samsonov A, Hurley SA, Dean DC, Field A, Alexander AL. MPnRAGE: A technique to simultaneously acquire hundreds of differently contrasted MPRAGE images with applications to quantitative T1mapping. Magnetic Resonance in Medicine. 2016;75(3). doi: 10.1002/mrm.25674 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Keenan KE, Stupic KF, Boss MA, Russek SE, Chenevert TL, Prasad PV, et al. Multi-site, multi-vendor comparison of T1 measurement using ISMRM/NIST system phantom. In Proceedings of the 24th Annual Meeting of ISMRM, Singapore, 2016. Abstract 3290 [Google Scholar]
  • 29.Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE transactions on medical imaging. 2001;20(1):45–57. doi: 10.1109/42.906424 [DOI] [PubMed] [Google Scholar]
  • 30.Patenaude B, Smith SM, Kennedy DN, Jenkinson M. A Bayesian model of shape and appearance for subcortical brain segmentation. NeuroImage. 2011;56(3):907–922. doi: 10.1016/j.neuroimage.2011.02.046 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface reconstruction. NeuroImage. 1999;9(2):179–194. doi: 10.1006/nimg.1998.0395 [DOI] [PubMed] [Google Scholar]
  • 32.Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, et al. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron. 2002;33(3):341–355. [DOI] [PubMed] [Google Scholar]
  • 33.Weiskopf N, Suckling J, Williams G, Correia M, Inkster B, Tait R, et al. Quantitative multi-parameter mapping of R1, PD(*), MT, and R2(*) at 3T: a multi-center validation. Frontiers in neuroscience. 2013;7:95. doi: 10.3389/fnins.2013.00095 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Fujita S, Hagiwara A, Hori M, Warntjes M, Kamagata K, Fukunaga I, et al. Three-dimensional high-resolution simultaneous quantitative mapping of the whole brain with 3D-QALAS: An accuracy and repeatability study. Magnetic Resonance Imaging. 2019;63:235–243. doi: 10.1016/j.mri.2019.08.031 [DOI] [PubMed] [Google Scholar]
  • 35.Schoemaker D, Buss C, Head K, Sandman CA, Davis EP, Chakravarty MM, et al. Hippocampus and amygdala volumes from magnetic resonance images in children: Assessing accuracy of FreeSurfer and FSL against manual segmentation. Neuroimage. 2016;129:1–14. doi: 10.1016/j.neuroimage.2016.01.038 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supp Video 1

Supporting Information Video S1: This animated GIF shows six examples of T1-weighted images from MPnRAGE reconstructed with and without motion correction. The top row shows two cases with mild motion artifacts, the middle row shows two cases with moderate motion artifacts, and the bottom row shows two cases with severe motion artifacts. In all cases retrospective motion correction improves visual sharpness of images.

NIHMS1587567-supplement-Supp_Video_1.gif (399.5KB, gif)
Supp Video 2

Supporting Information Video S2: This animated GIF shows six examples of quantitative T1-relaxometry images from MPnRAGE reconstructed with and without motion correction. The top row shows two cases with mild motion artifacts, the middle row shows two cases with moderate motion artifacts, and the bottom row shows two cases with severe motion artifacts. In all cases retrospective motion correction improves visual sharpness of images. These are the same cases and slices from Supporting Information Video S1.

NIHMS1587567-supplement-Supp_Video_2.gif (164.5KB, gif)
Supplement

Supporting Information Table S1: Mean T1 values and coefficients of variation (COV) for region of interest measurements from FSL FAST. Compared to the values without motion correction, the COVs with motion correction are on average 74% +/− 7% less. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/3 = 0.0167 are highlighted in gray.

Supporting Information Table S2: The coefficient of variation (COV) for region of interest measurements of T1 from FSL FIRST. Compared to the values without motion correction, the COVs after motion correction are on average 32% +/− 21% less. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Table S3: The coefficient of variation (COV) for region of interest measurements of T1 from FreeSurfer. Compared to the values without motion correction, the COVs with motion correction are on average 33% +/− 26% less. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Table S4: The mean T1 values for region of interest measurements from FSL FIRST. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Table S5: The mean T1 values for region of interest measurements from FreeSurfer. The p-values for the t-tests and f-tests are uncorrected before multiple comparisons. Statistically significant values after Bonferroni correction, i.e. p<0.05/15 = 0.0033 are highlighted in gray.

Supporting Information Figure S1: A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FAST. Shown are all measurements from all subjects and trials, color coded by subject. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

Supporting Information Figure S2: A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FIRST. Shown are all measurements from all subjects and trials, color coded by subject. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

Supporting Information Figure S3: A comparison of the quantitative T1 values measured in automatically segmented regions of interest using FreeSurfer. Shown are all measurements from all subjects and trials, color coded by subject. The y-axis shows the difference of T1 values after motion correction, normalized by the mean value, indicating an average increase of T1 after motion correction was applied. The black dashed line is the mean of the normalized difference, while the thin red lines are at +/− 1.96 standard deviations from the mean.

NIHMS1587567-supplement-Supplement.docx (512.4KB, docx)

RESOURCES