PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
J Magn Reson Imaging. Author manuscript; available in PMC 2010 September 1.
Published in final edited form as:
PMCID: PMC2753510
NIHMSID: NIHMS136077

Modulated Repetition Time Look-Locker (MORTLL): A Method for Rapid High Resolution Three-Dimensional T1 Mapping

Neville D. Gai, PhDcorresponding author§ and John A. Butman, MD, PhD

Abstract

Purpose

To demonstrate a modification of the Look-Locker (LL) technique that enables rapid high resolution T1 mapping over the physiologic range of intracranial T1 values, ranging from white matter to cerebrospinal fluid (CSF). This is achieved by use of a three-dimensional (3D) balanced steady-state free precession (b-SSFP) acquisition (for high signal-to-noise and resolution) along with variable repetition time to allow effective full recovery of longitudinal magnetization.

Materials and Methods

Two modifications to the Look-Locker technique were made to realize high resolution imaging in a clinically reasonable scan time. 3D balanced SSFP acquisition after an initial inversion pulse was followed by a variable repetition time. This technique makes it possible to image a volume of thin contiguous slices with high resolution and accuracy using a simple fitting procedure and is particularly useful for imaging long T1 species such as CSF. The total scan time is directly proportional to the number of slices to be acquired. The scan time was reduced by almost half when the repetition time was modified using a pre-designed smooth function. Phantoms and volunteers were imaged at different resolutions on a 3T scanner. Results were compared with other accepted techniques.

Results

T1 values in the brain corresponded well with full repetition time imaging as well as inversion recovery spin echo imaging. T1 values for white matter, gray matter and CSF were measured to be 755±10ms, 1202±9ms and 4482±71ms, respectively. Scan times were reduced by about half over full repetition time measurements.

Conclusion

High resolution T1 maps can be obtained rapidly and with a relatively simple post-processing method. The technique is particularly well suited for long T1 species. For example, changes in the composition of proteins in CSF are linked to various pathologies. The T1 values showed excellent agreement with values obtained from inversion recovery spin-echo imaging.

Keywords: T1 Map, Look-Locker, 3D, variable repetition time

INTRODUCTION

Fast and accurate T1 measurements are desired in a variety of applications. Various techniques have been proposed for acquiring T1 maps in-vivo. Most are based on one of two generic methods: (a) excitation using variable flip angles and (b) sampling of the saturation or inversion recovery curve.

The variable flip angle method can be performed relatively rapidly with at least two 3D fast field echo acquisitions, each with a different flip angle. This scheme has been applied in the head (1) and in body applications (2). However, it is well recognized that B1 inhomogeneity results in significant errors due to variation of the actual flip angles from the prescribed flip angle. Thus accurate T1 mapping with this technique requires a separate measurement to generate a field map to correct for the B1 inhomogeneity, particularly at higher field strengths. Furthermore, the range of T1 values which may be sampled is limited by the choice of flip angles and repetition times (1).

Ideally, sampling of the inversion recovery curve is obtained by repeatedly performing a single-echo spin-echo experiment with multiple inversion times and with a TR~5 times the maximum T1 value to be measured. This may be considered the gold standard for computing T1 in vivo (3). This method is not affected by B1 inhomogeneity but cannot be practically used for high resolution measurements due to long scan times. Saturation recovery reduces this scan time at the cost of reduced dynamic range.

In order to calculate T1 more rapidly, it is desirable to sample the inversion recovery curve repeatedly following a single inversion, rather than once as in a single-echo spin-echo experiment. To do so accurately, the IR curve should be disturbed minimally by the sampling. Various methods employed include turbo spin echo (4), sampling the stimulated echo (5), fast gradient echo or low angle shot (FLASH) (6), echo planar ((7), (8)) and balanced SSFP imaging ((9), (10), (11)). While SE and TSE typically require longer scan times for similar slice coverage and resolution than the other three techniques, echo planar imaging suffers from distortion and other phase related errors especially if the resolution required is quite high. These artifacts can be partially corrected with added scans and/or post-processing. Turbo field echo imaging alters the recovery curve substantially if the flip angle is higher than a few degrees (6).

Relatively, balanced SSFP imaging exhibits the best characteristics conducive to extended data acquisition with minimal perturbation of the recovery curve. A prior work (9) illustrated this property and showed that the recovery curve was barely perturbed even for flip angles as high as 50°. However, this is true only when T1~T2 or for small flip angles. A subsequent work (10) addressed the error resulting from the steady-state value being different from the initial value by applying a three parameter model. A drawback though is that about six to ten Look-Locker (LL) phases are needed for a good fit. This compromises either the scan time or the resolution or both. At larger flip angles, the off-resonance response of b-SSFP also suffers and results in inaccuracies in the determined T1 values. In addition, the three parameter fit with correction may result in some inaccuracies in T1 values even when flip angles are modestly larger (~25°). This is partly due to the correction being dependent on the relatively higher flip angle used which in turn makes it sensitive to B1 inhomogeneity at higher fields.

The above b-SSFP based methods used 2D multi-slice acquisition techniques. For thin contiguous slices acquired with high resolution, 2D multi-slice methods suffer from a variety of drawbacks. In addition to poor SNR and cross-talk, the variation in flip angle along the slice direction (due to B1 inhomogeneity) may result in inaccuracies. The deviation from the ideal recovery when sampled with b-SSFP depends on the flip angle, T1 and T2 values of the tissue. For tissues with long T1 and T2s (like CSF), this deviation is minimal even for large flip angles. However, for most other tissues (including white matter and gray matter), inaccuracies result from use of larger flip angles and may not be accurately corrected for with the three parameter model. As a result, it is preferable to image at lower flip angles for which 3D acquisition is most suitable. In addition, CSF bulk motion and inflow effects are minimized with 3D as compared to 2D acquisitions.

To accurately measure the T1 of intracranial fluids such as CSF, endolymph, or the vitreous humor, full recovery of CSF longitudinal magnetization is preferred which can take 18–20s, leading to very long scan times. The 3D b-SSFP acquisition has an effective scan time of one repetition time per slice as one z (slice) encoding step is followed by a single shot acquisition of phase encoding (y) steps. The scan time then increases with increasing number of acquired slices. To reduce this scan time, TR was modulated from TRmin to TRmax from shot to shot using a predefined smooth function. Just as an effective inversion time or an effective echo time is defined to the center of ky space, so also an effective repetition time can be defined for the center of kz space. Thus the center of k-space sees a TR corresponding to full recovery of the inverted magnetization. Survey of existing literature pointed to a previous description of a generic method based on variation of parameters including repetition time (12). T1 values in phantoms as well as T1 maps in normal volunteers were obtained using these two modifications of the Look-Locker sequence, namely 3D b-SSFP acquisition using a single shot for each slice encoding step and using a modulated TR. The implications of modulation of TR on magnetization along kz are analyzed. T1 maps obtained using the full TR are compared to those obtained using modulated TR per slice encoding. In addition, T1 maps obtained with our technique (labeled MORTLL for Modulated Repetition Time Look-Locker) are compared with a eight phase three parameter model (referred to as CORTLL for Constant Repetition Time Look- Locker) as described in (10) and (11) as well as to maps obtained using a single echo spin-echo sampling of the inversion recovery curve.

MATERIALS AND METHODS

Pulse Sequence

Previously, segmented Look-Locker acquisition techniques have been used in the multi-slice 2D context. The modified acquisition collects data for a 3D slab such that each inversion pulse is followed by a 3D slice encoded (kz step) single-shot (all ky phases) acquisition. Acquisition in a single repetition time is shown schematically in Figure 1. The figure shows the case when three Look-Locker phases for a slice encoding step are acquired. This is followed by a varying dead time during which the Mz recovers according to the dead time used. Such a scheme allows for control over the total scan time based on the number of slices desired. The number of LL phases of the slice encoded acquisition has a relationship with the resolution desired. Higher resolution acquisitions result in longer inversion times (TI1,2,3). The number of LL phases acquired in this case would be limited to a maximum of three or four. To compute the T1 of CSF (~4 sec), full recovery of the magnetization in the brain can be ~20 sec. However, with 3D acquisition, this recovery time can be effectively achieved for all slices provided the encoded slices near the center of k-space experience full longitudinal magnetization recovery. The dead time between acquisition of the LL phases and full recovery of Mz can be considerable. For example, for an image with 194 phase encodings (partial Fourier encoding for a 256 resolution image) with TR=3.8msec, shot duration is approximately 730msec. If three LL phases are acquired, the dead time would be more than 15 sec for a TR of 18 sec. So typically in the brain, scan time would be in excess of 9 minutes for about 25 slices with oversampling along z. Since most of the energy of the object is concentrated around kz=0 encoding, considerable reduction in total scan time can be obtained by varying the dead time from 0 (at kz = −kz,max) to the maximum value (at kz=0) and back to 0 (at kz= +kz,max) in a smooth fashion. The shot repetition time then goes from TRmin → TRmax → TRmin. A Blackman-Harris (BH) window is employed for the purpose so that the modulation in image space shows a response more faithful to the full recovery scheme. Blackman-Harris is a good general purpose window that is easy to implement and shows a favorable response in the transformed (image) domain.

Figure 1
Acquisition scheme for 3D IR-bSSFP based scheme. For each kz encoding, data are acquired as shown above. All ky lines for that slice encoding are acquired as a single-shot which is repeated 3 times to give 3 encoded images at the same kz location. The ...

A four term Blackman-Harris window is given by

w(t)=0.358750.48829cos(2πtT)+0.14128cos(4πtT)0.01168cos(6πtT)

The dead time is calculated based on the above window in the following fashion:

  1. Let M be the total number of encoding steps along kz.
  2. Calculate w(n) for T=M/2 and n=0, 1, 2, …M/2-1.
  3. Normalize step size using Δw(n) = w(n)/Σw(n) for n=0, 1,…,M/2-1
  4. Now calculate dead-time duration using Δtdd(n) = (TRmax-TRmin)×Δw(n). The dead times for the other half of the kz acquisitions then mirror those for n=0,…., M/2-1.

Simulations

Figure 2 shows the modulation effect of varying TR and 3D-bSSFP acquisition for different cases. Figure 2(a) shows the TR variation across slice encodings in k-space. Figure 2(b) shows the effect of TR modulation on a constant slab profile. The T1 is assumed to be 4 seconds while the number of slices is 64. The modulated waveform is obtained by multiplying the Fourier transform of the slab with a Hamming window (to reduce ringing effects) and the longitudinal magnetization value at each encoding step followed by an inverse transform. Figure 2(c) and 2(d) show the modulation effect of three phase b-SSFP acquisition and TR variation on measured pixel values along the z-axis. The modulation effect of a varying TR is most pronounced for long T1 species and high spatial frequency content (Figure 2c). A pixel with such characteristics along z (fixed (x,y) location) was chosen from reconstructed images acquired at a TI as seen from the slab profile. Values along z included pixels from CSF as well as WM although a worst case T1 value of 4 sec was used to determine the effect of modulated TR. Four cases were considered: (a) “Pure”: denoting pixel values that undergo minimal modulation when TR~5×T1 (b) modulation effect due to three phase LL acquisition with TR~5×T1 (c) modulation effect due to a shorter TR for which scan time is equivalent to MORTLL acquisition (~50% reduction) and (d) modulation due to three phase LL and modulated TR. Acquisition time was assumed to be 2500ms for (b), (c) and (d). The weighted mean error (WME) between the values obtained with constant TR and the value obtained with a modulated TR was 5.9% while χ2=1.74. A more typical case is when the variation along z is relatively smooth. One such pixel location was also chosen and the slab profile reconstructed using constant and varying TRs. Figure 2(d) shows that the modulation effect is minimal with the scheme. The weighted mean error between constant TR (b) and modulated TR (d) was a mere 1.06% while χ2=0.048. For both high frequency and low frequency data along z, the Blackman-Harris modulated TR scheme shows a better response than the case when the TR is constant at 9 sec (resulting in a similar total scan time). Table 1 shows the WME and χ2 values between (a) and (b), (c) and (d) for both data sets. (Note that χ2<23.7 indicates that the two sets being compared are not significantly different (significance level =0.05).)

Figure 2
(a): Variation scheme for the repetition time based on a Blackman-Harris window (top left); 2(b): Slab profile resulting from the three repetition time schemes (assuming 64 slices; top right): (1) Constant TR=18 sec (full recovery) (2) constant TR=5s ...
Table 1
Weighted mean error and χ2 values for ideal vs modulated data sets.

Simulations using Bloch equations (13) for Mz recovery with continuous b-SSFP acquisition were also done. The recovery curves for WM, GM and CSF for different flip angles are shown in Figure 3. The following values were used for T1 and T2 at 3T: T1(WM) = 1089msec, T2(WM) = 69msec; T1(GM) = 1820msec, T2(GM) = 99msec; ((14)); T1(CSF) = 4680msec, T2(CSF) = 2000msec; ( (15) and (16)). As can be seen, the error between the ideal recovery curve and the Mz sampled with the b-SSFP train increases with decreasing values of relaxation time and increasing flip angles. This was also shown in references (9) and (10). One would then expect the error to be relatively greater in white matter and least in CSF. Using values at typical TIs {300, 1050, 1800}ms from the modulated inversion recovery curve results in T1 values which are different from the ideal values by 0.46%, 0.44% and 0% for WM, GM and CSF respectively.

Figure 3
(a): Simulation results showing the effect of b-SSFP sampling with different excitation angles on the inversion recovery curve for white matter (left); (b) for gray matter (center) and (c) for CSF (right).

Imaging

All scans were performed on a Philips 3T Achieva scanner equipped with QUASAR dual mode gradients capable of maximum gradient amplitude of 80mT/m and a slew rate of 200T/m/s. Modifications to the sequence software were done for software release version 2.5.3.

Two phantoms with a T1/T2 ≈ 960/700msec and 2850/1150msec were imaged using the schemes described below. In addition, six volunteers (two females and four males ages 23 to 54yrs) were imaged near the basal ganglia under an IRB approved institutional protocol. After the nature of the procedure had been explained, written informed consent was obtained from all volunteers. Acquisitions with two different resolutions were done with our scheme. Scans with the following parameters were carried out for all volunteers:

Medium Resolution MORTLL (α=10°): FOV=23cm, TI1,2,3 ≈ {300, 1050, 1800}msec; variable TR with TRmin≈2.5sec and TRmax=18s, b-SSFP acquisition, 256×194 matrix, partial echo, partial Fourier encoding scan, resolution≈0.9×0.9×1mm; b-SSFP TR[TE=3.7/1.59msec, α=10°, #of slices=25, scan time≈5mins 20s

High Resolution MORTLL (α=10°): FOV=23cm, TI1,2,3 ≈ {550, 1600, 2650}msec, variable TR with shot TRmin≈3.8sec and shot TRmax=18sec, b-SSFP acquisition, 432×324 matrix, partial echo, partial Fourier encoding scan, resolution≈0.5×0.5×1mm b-SSFP TR[TE=3.9/1.7msec, , # of slices=25, scan time≈5mins 20s.

Low Resolution 8 phase CORTLL (α=10°): FOV=23cm, 8 Look-Locker phases with TI1,2,3,4,5,6,7,8 ≈ {135, 450, 765, 1080, 1395, 1710, 2025, 2340}msec, constant shot TR=18s, 128×94 matrix, partial echo, partial Fourier encoding scan, resolution≈1.8×1.8×1 m m b-SSFP TR[TE=3.7/1.59msec, α=10°, #of slices=25, scan time≈5mins 20s. Lower resolution was used due to the increased number of LL phases.

Note that the exact inversion times and shot durations vary slightly due to factors such as slab angle and patient weight.

For comparison purposes, the following scans were carried out for two volunteers:

IR Spin Echo: FOV=23cm, pFOV=0.8, TR=10s, TE=10msec, TI1,2,3={800, 1200, 1800}msec, 80×64 scan matrix, resolution ≈2.9×2.9×2mm scan time 8:50s. (The TR time was lower to keep scan times reasonable. Comparison for only WM and GM was done with the SE based T1 map.)

8 phase CORTLL with α=25°: To test 3 parameter model with higher flip angle.

Medium Resolution MORTLL with α=25°: To test the observation from simulation showing greater deviation from the ideal value.

Medium Resolution 3 phase CORTLL with α=10° and shot TR=18 sec. The total scan time in this case was 9 mins 40 sec. Note that MORTLL results in a 45% savings in scan time over the equivalent CORTLL scan.

In each case, a b-SSFP single-shot was prepared using a (α/2—TR/2) pulse and ten dummy (α—TR) acquisitions prior to data collection to help establish steady-state and reduce artifacts. Use of linear encoding along ky furthermore ensures that signal fluctuations are diminished before the ky=0 line is sampled. Sampling along kz (slice encoding) was also linear.

Analysis

On observing the magnetization recovery curve (Figure 3), one notes that the recovery can be approximated using the simple two parameter (M0(x,y){1–2e–TI/T1}) model provided the flip angle is kept low (~10°). We fit our acquired 3 LL-phase data to this model. This simplifies curve fitting and in conjunction with linear encoding for the relatively long single-shot phase minimizes the need for steady-state and longitudinal magnetization modulation based correction (as used in (10) and (11)). A three parameter model (ABe–TI/T1) along with correction for steady-state signal (which perturbs the recovery signal) was used for the low resolution eight LL phase CORTLL acquisition for comparison with prior literature. The Nelder-Mead (simplex) method was used to fit the data to both exponential models.

T1 measurements for WM, GM and CSF were made by drawing ROIs in white matter outside the caudate nucleus, inside the caudate nucleus (GM) and in the ventricles (CSF). All ROIs were placed in the same areas for all volunteers.

RESULTS

Phantom experiments

Figure 4 shows T1 maps for the two phantoms (T1≈2850ms and 950ms) obtained with intermediate (medium resolution MORTLL) and high resolution MORTLL 3D three LL phase two parameter model, 8 LL phase 3 parameter model and IR-spin echo. The measured values in a ROI are shown in Table 2. All T1 values including the ones measured with a relatively larger flip angle of 25° show excellent agreement with the one obtained with IR-SE measurements. The 8 phase LL with a flip of 10° shows the largest error of 1.3% compared with IR-SE.

Figure 4
T1 maps obtained with (a) IR spin-echo (b) 3 LL phase two parameter MORTLL with resolution ~0.9×0.9×1mm (c) 3 LL phase two parameter MORTLL with resolution ~0.5×0.5×1mm (d) 8 phase three parameter CORTLL (α=10°). ...
Table 2
T1 values (in msec) in two phantoms using different acquisition schemes.

For the lower T1/T2 value phantom, both the 8 phase LL acquisitions (α=10° and α=25°) using the three parameter model and correction show significantly lower T1 value (~8.6%) when compared with IR-SE.

In-vivo Real Images from the 3 phase MORTLL experiments

Figure 5(a) shows the three reconstructed real images acquired in the brain using MORTLL for determining the T1 maps. Figure 5(b) shows the corresponding T1 map and 5(c) the corresponding proton density map. The proton density map reflects B1 inhomogeneity while the T1 map is relatively artifact free. T1 calculation for one 512×512 slice using our simple two parameter model takes about 10 mins on a laptop with a Intel Pentium M processor (1.73GHz) and 2GB of RAM while it takes about 45 sec on a dual CPU Intel Xeon (3.2GHz) machine.

Figure 5
(a) Three images obtained at TI=[280, 1016, 1751]msec (W/L : 550/−25) (b) T1 map and (c) corresponding M0(x,y) map derived from the three images using the two parameter model. M0(x,y) map clearly shows B1 inhomogeneity while the T1 map is free ...

Comparison with Full Repetition Time T1 Maps

For two volunteers, maps obtained with medium resolution MORTLL were compared with those obtained with a constant TR. The images were found to differ by −1.0%, 0.95% and −0.5% in WM, GM and CSF, respectively. The accuracy of MORTLL is not much different from CORTLL despite the substantial savings in scan time. Using a higher flip angle (α=25° instead of α=10°) with the MORTLL resulted in WM, GM and CSF values which were different by 9.2%, −0.6% and −2.25%, respectively. This follows theory in that the error in computed values increases with increasing flip angles for short T1 (WM) species.

Comparison of 3 phase MORTLL with 8 phase CORTLL and IR-SE

Figure 6 shows T1 maps for a slice obtained in the basal ganglia region with 3 phase MORTLL, 8 phase low resolution CORTLL and inversion recovery spin echo using parameters described in the Methods section. Figure 7 shows the reformatted sagittal and coronal views of the reconstructed T1 maps for the 25 slices. A bar graph of the measured values in GM, WM and CSF across the six volunteers for medium and high resolution MORTLL and low resolution CORTLL is shown in Figure 8. Medium resolution and high resolution 3 phase LL T1 values show no statistically significant differences between WM (χ2=5.1) and GM (χ2=5.5) and only a marginally significant difference in CSF (χ2=15.3). On the other hand, medium resolution 3 phase LL T1 values show substantially significant differences from values obtained with the 3 parameter 8 phase LL values in WM, GM and CSF (χ2=55, 48 and 182), respectively. (The χ2 value needs to be less than 11.1 for a significance level of 0.05.) A one-way ANOVA test between the three scans reveals statistically significant differences in all three tissues (F=23.17, 7.75 and 26.2 > 3.68 for WM, GM and CSF). A paired t-test shows no statistically significant difference between medium resolution MORTLL and high resolution MORTLL (t-value=1.63, 0.86, 0.133 <1.812 for WM, GM and CSF). However, medium resolution MORTLL showed statistically significant differences with 8 phase CORTLL (t-value = 4.6, 2.71 and 5.7 > 1.812 for WM, GM and CSF, respectively). A statistical significance level of 0.05 was used for all comparisons.

Figure 6
T1 maps in mid-brain obtained with (a) 3 phase two parameter MORTLL with resolution ~0.9×0.9×1mm (b) 3 phase two parameter MORTLL with resolution ~ 0.5×0.5×1mm (c) low resolution (1.8×1.8×1mm) 8 LL phase ...
Figure 7
Reformatted T1 maps: (a) sagittal and (b) coronal views for the 25 slice slab.
Figure 8
Bar graph shows T1 values measured in WM, GM and CSF for the six volunteers using three different scanning schemes: (a) Medium res. 3 phase MORTLL (b) High res. MORTLL (c) 8 phase CORTLL with correction. (Flip angle is 10°.)

Figure 9 compares the T1 values obtained with medium resolution MORTLL (α=10°) to 8 phase CORTLL with two different excitation angles (α=10°, 25°) and to inversion recovery spin echo. The two techniques (3 phase 2 parameter LL and 8 phase 3 parameter LL) are compared against values obtained with IR-spin echo for two of the volunteers. The differences are shown in Table 3. T1 for CSF was 7.8% lower for 8 phase CORTLL when compared with medium resolution MORTLL.

Figure 9
T1 values for WM, GM, CSF measured for four different techniques: (a) IR-SE (b) Med-res. 3 phase MORTLL (c) Low res. 8 phase CORTLL (α=10°) (d) Low res. 8 phase CORTLL (α=25°). Note that CSF measurements with IR-SE were ...
Table 3
Variation in T1 values between IR-SE and 3 phase 2 parameter LL (MORTLL) and 8 phase 3 parameter LL (8ph CORTLL).

DISCUSSION

Partial Fourier encoding (ky) as well as partial echo (along kx) was used to keep shot durations as well as individual b-SSFP TRs short and hence increase temporal resolution (LL phases) and decrease off-resonance effects. Use of low flip angles also improves off-resonance response. Parallel imaging along the phase encoding direction could be used to reduce effective inversion times by reducing shot duration as long as steady-state considerations are kept in mind. SENSE along the slice encoding direction will reduce the total scan time by the SENSE factor used. Preliminary results with using SENSE along the slice encoding direction (with a favorable sagittal prescription) showed slightly elevated T1 values on phantoms and comparable values in GM, WM and CSF to T1 maps obtained without SENSE. Figure 10 shows a slice obtained with a slice reduction factor of 2; as expected, the scan time is almost half that for a corresponding MORTLL sequence without SENSE.

Figure 10
T1 map obtained using SENSE factor 2 along the z direction with 3 phase MORTLL. This results in a scan time reduction.

Another observation from the simulations is that for small flip angles the recovery curve follows the ideal recovery closely near the inversion pulse and diverges away from the inversion pulse. Hence, one would expect the error to increase as the later LL-phases occur further away from the inversion pulse. This is the case for higher resolution scans and WM imaging. CSF values are relatively unaffected with the error being about 0.4% when the resolution is about 0.5mm (432 readout points) and with a flip angle of 25°. On the other hand, the error between the ideal T1 value and the one derived from the modulated IR curve is about 10% in WM at a flip angle of 25° for the same resolution. As reflected in the Results section, correction for modulation and steady-state effects based on prior 2D b-SSFP sampling strategies (10) with larger flip angles still results in inaccurate values for T1 in WM, GM and to a lesser extent in CSF. With correction done as given in (11), the values showed even higher divergence from those obtained with IR-SE and our scheme. For accurate 3D T1 mapping of WM and GM, the maximum flip angle should be constrained to low angles as done in this work. Our results show excellent correlation with spin-echo based T1 mapping for GM and WM. In addition, being a Look-Locker based technique, it is more robust to motion related artifacts since acquisitions for the different inversion times are all obtained in a single TR.

Although the T1 values for WM and GM used in simulations (based on values reported elsewhere) were higher than the values calculated here, Wansapura et al. (17) have reported T1 values for WM and GM that more in line with the values reported here. WM T1 was around ~ 830ms while GM values were found to vary between 1276–1392ms depending on the area of the brain.

To determine the efficacy of the variable TR scheme for fewer slices, the values obtained with a much lower slice encoding scheme of 13 slices were compared with those obtained from medium resolution MORTLL with 25 slices. The values differ by 1.3%, −3.3% and −2.3% for WM, GM and CSF. It is apparent from the variable TR scheme that errors are likely to increase as the number of slice encodings is reduced even further. Higher resolution scans (voxel size<0.5×0.5×0.5mm) can be obtained in regions of high signal such as CSF spaces with use of higher flip angles. However, T1 values in white matter get noisier and less accurate at such resolutions.

The technique can be easily adapted for low-high encoding scheme where either kz or ky or both are acquired in a centric fashion with lower frequencies being acquired first. However, in this case, the number of dummy pulses might need to be increased considerably to avoid initial signal fluctuation related inaccuracies and artifacts. Note that the repetition time modulation is symmetric about the center of k-space implying that symmetric sampling along the modulation direction is needed. This typically implies full k-space sampling along that direction. Full k-space sampling is more likely along the slice encoding direction (kz) since typically phase encoding (ky) may use partial Fourier encoding.

Note that the total scan time for our scans is relatively high (5mins+) for 25 slices since we map CSF T1 values. If only GM and WM mapping is desired, the effective TR can be reduced considerably (~7s) resulting in a scan time of around 1min 45sec for 25 slices with an in-plane resolution of 0.5×0.5mm.

No additional scans for excitation angle correction are required as with multiple flip angle based schemes. With the multiple flip angle method, for a range of tissue T1 values, the optimal choice for TR and flip angles can still provide poor discrimination between tissues with different T1, affect the SNR of the measurements and also the accuracy of the values. As noted in (1), the error in calculated T1 values increases with higher T1 values. This observation is further reinforced in other works (18) where the reported mean value for CSF using variable flip angles (and applying correction for B1 inhomogeneity) was 2180msec at 3T which is grossly lower than those reported with inversion recovery methods which range from 3700msec (7) to 5130msec (8). Although inversion recovery based measurements should provide accurate mapping for longer T1 values, there is still a fair amount of discrepancy between techniques, most likely a result of applied corrections for flip angle, repetition time, acquisition technique or the use of a three parameter fitting model. In addition, as the repetition time between inversion pulses is reduced, saturation effects impact calculations for long T1 values. In particular, T1 values from the eight phase LL (three parameter model) with correction seem to compare less favorably to IR-SE values than values obtained by the three phase two parameter model used here. As indicated elsewhere, this could be partly due to the dependency of the correction on the excitation angle (and hence B1 inhomogeneity). In addition, parameters derived from curve fitting are used to correct for steady-state related effects. This could further introduce errors as a three parameter model is inherently less accurate and also shows a strong dependency on sampling times (19). Non steady-state effects would be more pronounced for the 6–10 LL phase experiments since the duration of each phase is relatively shorter. Shorter phases imply greater non steady-state effects when compared with longer phase durations with linear sampling.

For the scans performed, the specific absorption rate (SAR) and peripheral nerve stimulation (PNS) were always within FDA approved limits. A worst case estimate for all system limits and power deposition was derived by calculating the values based on the minimum sequence time (i.e. with dead time=0). Flow and motion (gross motion, CSF pulsatility etc) can result in errors in T1 calculations especially at higher resolutions as the difference between inversion times increases. Some fine line artifacts and blooming could be seen in certain slices for high resolution MORTLL images around areas of high susceptibility (typically near air tissue interface areas like the bottom of the frontal lobe in a sagittal view). However, most of this artifact energy was confined to proton density maps leaving T1 maps relatively free of them.

In conclusion, a simple two parameter model can be used to fit high spatial resolution data with a minimum of three Look-Locker phases as long as the excitation angle is kept low (~10°). Variable repetition time was used to reduce total acquisition time. T1 maps obtained with such a technique were compared with inversion recovery single-echo spin-echo based maps and showed excellent agreement.

ACKNOWLEDGEMENTS

This work was supported by the Intramural Research Program of the NIH Clinical Center.

This research was supported by the Intramural Research Program of the NIH Clinical Center.

REFERENCES

1. Deoni SC. High-resolution T1 mapping of the brain at 3T with driven equilibrium single pulse observation of T1 with high-speed incorporation of RF field inhomogeneities (DESPOT1-HIFI) J Magn Reson Imaging. 2007;26:1106–1111. [PubMed]
2. Treier R, Steingoetter A, Fried M, Schwizer W, Boesiger P. Optimized and combined T1 and B1 mapping technique for fast and accurate T1 quantification in contrast-enhanced abdominal MRI. Magn Reson Med. 2007;57:568–576. [PubMed]
3. Steen RG, Gronemeyer SA, Kingsley PB, Reddick WE, Langston S, Taylor JS. Precise and accurate measurement of proton T1 in human brain in vivo: Validation and preliminary clinical application. J Magn Reson. 1994;4:681–691. [PubMed]
4. Conturo RR, Price RR, Beth AH, et al. Improved determination of spin density, T1 and T2 from a three-parameter fit to multiple-delay-multiple-echo (MDME) NMR images. Phys Med Biol. 1986;31:1361–1380. [PubMed]
5. Haase A, Frahm J. NMR imaging of spin-lattice relaxation using stimulated echoes. J Magn Reson. 1985;65:481–490.
6. Deichmann R, Haase A. Quantification of T1 values by snapshot FLASH NMR imaging. J. Magn Reson. 1992;96:908–912.
7. Clare S, Jezzard P. Rapid T1 mapping using multislice echo planar imaging. Magn Reson Med. 2001;45:630–634. [PubMed]
8. Henderson E, McKinnon G, Lee TY, Rutt BK. A fast 3D Look-Locker method for volumetric T1 imaging. Magn Reson Imaging. 1999;17:1163–1171. [PubMed]
9. Scheffler K, Hennig J. T1 quantification with inversion recovery True-FISP. Magn Reson Med. 2001;45:720–723. [PubMed]
10. Schmitt P, Griswold MA, Jakob PA, et al. Inversion recovery TrueFISP: Quantification of T1, T2 and spin density. Magn Reson Med. 2004;51:661–667. [PubMed]
11. Messroghli DR, Radjenovic A, Kozerke S, et al. Modified Look-Locker inversion recovery (MOLLI) for high-resolution T1 mapping of the heart. Magn Reson Med. 2004;52:141–146. [PubMed]
12. Coleman JM, Kaufman L, Kramer DM. MRI using variable imaging parameter(s) within a single sequence. US Patent. 5239266 1993.
13. Hargreaves BA, Vasanawala SS, Pauly JM, Nishimura DG. Characterization and reduction of the transient response in steady-state MR imaging. Magn Reson Med. 2001;46:149–158. [PubMed]
14. Stanisz GJ, Odrobina EE, Pun J, et al. T1, T2 relaxation and magnetization transfer in tissue at Magn Reson Med. 2005;54:507–512. [PubMed]
15. Warntjes JBM, Dahlqvist O, Lundberg P. Novel method for rapid, simultaneous T1, T2*,and proton density quantification. Magn Reson Med. 2007;57:528–537. [PubMed]
16. Pell GS, Waites AB, Breillman RS, Jackson GD. Voxel based relaxometry. Proceedings of the 11th Annual Meeting of ISMRM; Toronto. 2003. abstract 136.
17. Wansapura JP, Holland SK, Dunn RS, Ball WS. NMR relaxation times in the human brain at 3.0 Tesla. J Magn Reson Imaging. 1999;9:531–538. [PubMed]
18. Wang J, Qiu M, Kim H, Constable RT. T1 measurements incorporating flip angle calibration and correction in vivo. J Magn Reson. 2006;182:283–292. [PubMed]
19. Zhang Y, Yeung H, O'Donnell M, Carson PL. Determination of sample time for T1 measurement. J Magn Reson Imaging. 1998;8:675–681. [PubMed]