Fig 1.
Schematic diagram of the ME-MP2RAGE sequence.
After an adiabatic inversion pulse (INV), two GRE readout blocks are collected with excitation pulse flip angles, α1 and α2. Both GRE blocks consist of n acquisitions of k-space lines, each of duration TR,GRE, stepping linearly through the second phase-encoding direction (as in standard MP2RAGE). To each TR,seq corresponds a k-space line acquisition for the first phase-encoding direction. The center of k-space is acquired at times TI,1 and TI,2. Examples of the additional mono-polar gradient lobes of the ME readouts are indicated by blue color. The sequence repetition time, TR,seq, is defined as the time between two successive inversion pulses. The total acquisition time is thus defined by TR,seq multiplied by the number of steps in the second phase-encoding direction. TA, TB, and TC denote, respectively, the durations from the initial inversion pulse to the onset of the first GRE block, between the two GRE blocks, and from the end of the second GRE block to the next inversion pulse. Note that the first and the second phase-encoding direction may be interchanged.
Fig 2.
Estimated T1 values as a function of (a) the ME-MP2RAGE and (b) the MP2RAGE signal intensity parameter ρ.
The green lines correspond to the effective acquisition parameters while red/blue lighter/darker lines indicate, respectively, ±20% and ±40% offsets of . The flip angle values are adjusted for the accuracy factor ηα. Note that the estimated T1 is limited to an upper value of approximately 3 s for the acquisition parameters that were used for MP2RAGE, which would result in T1 values exceeding this limit (e.g., in CSF) to be underestimated.
Table 1.
Details of the manually optimized acquisition parameters that were used for the simulations in Fig 2 and for the acquisitions of Study 2 with a nominal spatial resolution of 0.6 mm (isotropic).
Fig 3.
Simulations for the expected accuracy and precision ofT1 and maps at SNR = 50.
Rows refer to T1 (a, b) and (c, d) simulations, while columns indicate the choice of the simulation parameters reflecting the acquisition: ME-MP2RAGE (a, c), MP2RAGE (b) or ME-FLASH (d). Each panel contain a plot of the estimated relaxation time as a function of the exact value, with the error bars indicating the standard deviations, σT, for N = 20000 simulations. The distance from the identity line indicates the expected accuracy, while the size of the error bars reflect the expected precision. For T1 / exponential recovery simulations, the range 0.5–3.5 s was probed, and the exact acquisition parameters simulated were TI,(1,2) = 800, 2400 ms for ME-MP2RAGE, and TI,(1,2) = 750, 2900 ms for MP2RAGE. For T2 / exponential decay simulations, the range 2–60 ms was probed, and the exact acquisition parameters simulated were nE = 4, TE,1 = 2.5 ms, ΔTE ≈ 4.2 ms for ME-MP2RAGE, and nE = 5, TE,1 = 3.0 ms, ΔTE = 6.0 ms for ME-FLASH.
Fig 4.
inhomogeneity displayed as histograms of the flip-angle accuracy factor inside the brain of six healthy human volunteers.
Voxels outside the head or not containing brain tissue were masked out. The dashed lines indicate the mean (thick line) plus/minus the standard deviations (thin lines) of ηα across subjects.
Fig 5.
Example of the images obtained from an ME-MP2RAGE acquisition.
Each row represents a different echo time. The columns show in order: first inversion magnitude (1st) and phase (2nd); second inversion magnitude (3rd) and phase (4th). Magnitude images are shown in arb.units, while phase image are in radians, both using a gray scale. Note that: (i) the phase images for the first inversion point show an abrupt change in their value corresponding to the zero crossing of the signal in the T1 recovery curve; (ii) the phase images for the second inversion point present some coil combination pole artifacts resulting in a corresponding degradation of the QSM maps at these locations.
Fig 6.
Single subject acquisition of T1 maps acquired with (a) ME-MP2RAGE, (c) MP2RAGE and (e) their difference, and test-retest reproducibility evaluation with voxel-by-voxel correlation 2D histograms for (b) ME-MP2RAGE, (d) MP2RAGE, and (f) ME-MP2RAGE versus MP2RAGE.
Note how the CSF voxel are underestimated by the MP2RAGE as a result of the specific choice of the acquisition parameters.
Fig 7.
Single subject acquisition of maps acquired with (a) ME-MP2RAGE, (c) ME-FLASH and (e) their difference, and test-retest reproducibility evaluation with voxel-by-voxel correlation 2D histograms for (b) ME-MP2RAGE, (d) ME-FLASH, and (f) ME-MP2RAGE versus ME-FLASH.
Fig 8.
Single subject acquisition of χ maps acquired with (a) ME-MP2RAGE, (c) ME-FLASH and (e) their difference, and test-retest reproducibility evaluation with voxel-by-voxel correlation 2D histograms for (b) ME-MP2RAGE, (d) ME-FLASH, and (f) ME-MP2RAGE versus ME-FLASH.
Table 2.
Summary of the group averages μg and SDs σg results of the correlation parameters for the T1, and χ maps acquired during Study 2.
Table 3.
Group averages, μg, and SDs, σg, for the ROI-based analysis of ME-MP2RAGE acquisitions.
Abbreviations: WM = White Matter; Fro. = Frontal Lobe; Tem. = Temporal Lobe; Par. = Parietal Lobe; Occ. = Occipital Lobe; Ins. = Insula; Put. = Putamen; Caud. = Caudate; Tha. = Thalamus; Cereb. = Cerebellum. Susceptibility is indicated here by Δχ as a reminder of the values being referenced to an arbitrary offset, implying that the group average and SD values are expected to be biased by that and, possibly, by the pole artifacts of the phase images.
Fig 9.
2D correlation histograms showing the effects of registration.
Columns refer to T1 (a, c), (b, d) and χ (e, f) maps, while rows show either mis-registration (a, b, c) or self-registration (d, e, f) effects, respectively. The mis-registration and the self-registration effects are here illustrated by comparing each map with itself: after the application of a small transformation without registration (mis-registration) or after the application of a large transformation followed by a registration step (self-registration). The x-axis indicate the unmodified map, while the y-axis indicate the transformed or self-registered map. The small tranformation considered for the misregitration is a 0.5px translation followed by a 0.5° rotation. The large transformation considered for the self-registration is a 10° rotation.