Measuring Renal Tissue Relaxation Times at 7T

Abstract

As developments in RF coils and RF management strategies make performing ultrahigh field renal imaging feasible, understanding the relaxation times of the tissue becomes increasingly important for tissue characterization and sequence optimization. By using a magnetization prepared single breath-hold fast spin echo imaging method, human renal T₁ and T₂ imaging studies were successfully performed at 7T while addressing challenges of B1+ inhomogeneity and peak short-term specific absorption rate (SAR). At 7T, measured renal T₁ values for the renal cortex and medulla (mean ± S.D.) were 1661 ± 68 ms and 2094 ± 67 ms, and T₂ values were 108 ± 7 ms and 126 ± 6 ms. For comparison, similar measurements were made at 3T where renal cortex and medulla T₁ values of 1261 ± 86 ms and 1676 ± 94 ms, and T₂ values of 121 ± 5 ms and 138 ± 7 ms were obtained. Measurements at 3T and 7T were significantly different for both T₁ and T₂ values in both renal tissues. Reproducibility studies at 7T demonstrated that T₁ and T₂ estimations were robust with group mean percentage differences of less than 4%.

INTRODUCTION

Motivated by the promise of higher signal noise ratio (SNR), increased resolution and/or reduced imaging time, new or better tissue contrast and improved parallel imaging performance, human MRI at ultrahigh magnetic fields (≥ 7T) has been a major research focus (1–5) despite the numerous challenges (6). Recently, with advances in RF coil engineering, RF shimming strategies and acquisition methods (7–10), UHF MRI technical development and translational research has been expanded from the human brain (5,11–14) to the abdominal and pelvic organs, e.g. heart, prostate and kidneys (15–26).

With respect to the kidney, increased SNR and reduced imaging time has the potential to reduce motion-related artifacts and improve imaging quality. For techniques such as arterial spin labeling (ASL) perfusion imaging, increased longitudinal relaxation times (T₁) and SNR at 7T are particularly advantageous by enabling shorter acquisition times with the potential of high resolution, large organ coverage and in a single breath-hold (24–26).

Efficient, optimal and quantitative renal imaging requires the knowledge of tissue longitudinal (T₁) and transverse (T₂) relaxation times (27). Measuring renal T₁ and T₂ itself can also provide potentially valuable information for diagnosis and prognosis (28). For example, the potential of renal tissue T₁ to characterize and differentiate specific renal diseases states has been demonstrated (29–31). Renal tissue T₁ is also needed for quantifying renal blood flow in ASL imaging (32) and optimizing the inversion time in non-contrast enhanced renal angiography studies (33). As tissue relaxation times are field-dependent and cannot be accurately predicted by theoretical calculation, in vivo studies must be performed to estimate these relaxation parameters. However, to date, no human studies have been performed to estimate renal relaxation times at 7T.

Measuring renal T₁ and T₂ is challenging due to respiratory motion. Although respiratory triggering can help reduce the resulting artifacts, an inconsistent respiratory rate and variations in end expiratory position between breaths can result in significant variations in kidney position confounding accurate image co-registration and subsequent relaxation time estimations (30). Therefore, single breath-hold imaging approaches are preferred. The two major imaging methods that have been predominantly used for renal tissue relaxation time estimations within a single breath-hold are true fast imaging with steady state precession (TrueFISP) (34) and single shot fast spin echo (ss-FSE) (35). Compared to TrueFISP, ss-FSE has improved tolerance to B₀ inhomogeneity. As field inhomogeneity increases with field strength, ss-FSE is arguably the preferred method to use for imaging readout at 7T. However, measuring renal T₁ and T₂ at 7T faces new challenges, including B₁+ inhomogeneity, reduced RF efficiency and specific absorption rate (SAR) limits which are more rapidly reached.

In this work, we describe a method based on an ss-FSE sequence that can measure renal longitudinal and transverse relaxation times at 7T within a single breath-hold. The B₁+ inhomogeneity and RF efficiency at 7T were managed by using subject dependent B₁+ shimming (8) and calibration. Using this technique, measurements of renal T₁ and T₂ values were acquired at both 3T and 7T in healthy normal volunteers. The reproducibility of these measurements at 7T was determined in repeated studies.

MATERIALS AND METHODS

Subjects

A total seven healthy volunteers (five males: 47 ± 17 years and two females: 34 ± 18 years, mean ± S.D.) provided written informed consent prior to being studied according to a local IRB approved protocol. Of the seven, five participated in all three studies: relaxometry studies at 3T and 7T, and repeat measurements at 7T. The repeat measurements at 7T were performed to determine reproducibility and involved two sessions about one week apart (6–8 days, median 7 days).

MRI

3T MRI studies were performed on a Siemens TIM Trio MRI scanner (Malvern, Pennsylvania) with a 60 cm diameter horizontal magnet, a body coil for RF transmission, and a body surface array (two rows of 3 elements) anteriorly combined with the spine array (two rows of 3 elements) posteriorly for signal reception.

7T MRI studies were performed on a Siemens 7T whole body MRI scanner with 58 cm diameter horizontal bore, using a 16-channel transceiver TEM stripline array (10) driven by a series of sixteen 1 kW amplifiers (CPC, Pittsburgh, PA). The RF power deposition and reflection from each channel was monitored separately and continuously by using a power-monitoring system built in-house (17). The power limit set for each channel was based on simulation results and according to the IEC guidance (33).

Imaging Sequences

A magnetization prepared single-shot fast spin echo (ss-FSE) imaging sequence was used for relaxometry measurements, as shown in Figure 1. T₁ mapping was performed using an inversion recovery preparation with variable inversion times (IR-prep ss-FSE). To achieve uniform inversion across and within an imaging slice, the inversion was performed using a slice-selective adiabatic hyperbolic secant RF pulse with an inversion slab that was twice the imaging slice thickness. At 3T, the inversion was achieved with a 15.36 ms hyperbolic secant pulse with 3 kHz bandwidth (36); at 7T an HS4 pulse with 20 ms duration and 1 kHz bandwidth was used (37). T₂ mapping was performed by inserting a Carr-Purcell-Meiboom-Gill (CPMG) style refocusing pulse train (38,39) after the excitation and prior to ss-FSE readout, and varying number of echoes in steps of two (CMPG-prep ss-FSE). The same echo spacing was used for both CPMG-prep and the ss-FSE readout.

Figure 1.

T₁ and T₂ mapping sequence diagrams: (a) Sequence diagram for T₁ mapping using varied inversion times after an initial dummy scan. Inversion was accomplished by a hyperbolic secant (HS) RF pulse. (b) Sequence diagram for T₂ mapping using a Carr-Purcell-Meiboom-Gill style refocusing echo train (T_2CPMG) with varying numbers of refocusing echoes. To minimize the short-term specific absorption rate (SAR) at 7T, the ss-FSE readout employed the use of hyperechoes. Note: *ss-FSE readout refers to the data sampling for refocused echoes excluding the RF excitation.

With this approach, the T₁ and T₂ maps can each be acquired within a single breath-hold to minimize respiratory artifacts (35). At the beginning of both T₁ and T₂ acquisitions, a dummy scan was performed to establish spin steady state. The saturation recovery time (Ts) between successive measurements was kept constant to avoid variable T₁ weighting (Figure 1). To reduce short-term SAR at 7T, variable-rate selective excitation (VERSE) RF pulses (40) were used for excitation, and the hyperecho phase and amplitude modulation scheme was used for ss-FSE readout (41).

For 3T studies, a transverse imaging slice with a large field of view (FOV) was used to cover both kidneys. For 7T studies, a transverse imaging slice with a small FOV covering one side of the kidneys was used. Focusing on a signal kidney allowed greater B₁+ homogeneity with increased efficiency to minimize short-term SAR and to improve uniformity for the adiabatic inversion preparation.

3T Acquisition Parameters

For both T₁ and T₂ measurements, imaging parameters for the ss-FSE readout were: FOV = 360 × 360 mm², matrix size = 256 × 256, in-plane resolution = 1.4 × 1.4 mm², excitation/refocusing RF flip angles (FAs) = 90°/180°, slice thickness = 5 mm, anterior to posterior phase encoding direction, parallel imaging accelerator factor using GRAPPA = 2 with 24 separately acquired reference lines, partial Fourier = 4/8, total 7 measurements with the first one as a dummy scan. For T₁ measurements, the following parameters were used: TR/TE 3000/29 ms, bandwidth = 814 Hz/pixel and echo spacing = 4.12 ms, and inversion times = {100, 150, 300, 500, 800, 1200} ms; for T₂ measurements, TR/TE = 3000/35 ms, bandwidth = 476 Hz/pixel, echo spacing = 4.98 ms, and six different numbers of preparation echoes = {4, 8, 16, 24, 32, 40}. The total acquisition times were less than 23 s and 19 s for T₁ and T₂ measurements, respectively.

7T Acquisition Parameters

Imaging parameters for T₁ and T₂ measurements at 7 T were: TR = 3000–4000 ms, FOV = 192 × 192 mm², in-plane resolution = 1.5 × 1.5 mm², slice thickness = 5 mm, phase encoding direction = left-right or anterior-posterior with 50~80% oversampling, partial Fourier = 5/8, parallel imaging acceleration factor using GRAPPA = 4 with 24 separately acquired reference lines, and hyperecho flip angle = 90 degree. T₁ measurements used TE = 16 ms, 814 Hz/pixel bandwidth with 3.96 ms echo spacing. T₂ measurements used TE = 20 ms, 439 Hz/pixel bandwidth with 5.0 ms echo spacing. To further minimize short-term SAR for single breath-hold T₁ and T₂ imaging, the inversion times and the number of preparation echoes for six measurements were specifically arranged in the following order: {100, 1200, 150, 800, 300, 500} ms and {40, 4, 32, 8, 16, 24}, respectively. The maximal imaging acquisition times across all subjects were less than 29 s and 25 s for T₁ and T₂ measurements, respectively, with a 4 s repetition time.

B₀ and B₁+ optimization

On both 3T and 7T, local B₀ optimization was accomplished by acquiring volumetric phase maps within a single breath-hold (42).

For 7T imaging, B₁+ optimization was performed in the selected kidney region by using phase-only B₁+ shimming, followed by a flip angle calibration over the same region. B₁+ shimming was performed by acquiring a small flip-angle calibration scan and optimizing B₁+ using a tradeoff between homogeneity and efficiency (33). B₁+ maps were subsequently generated using a 2D implementation of the Actual Flip-angle Imaging (AFI) technique (43) acquired in a single breath hold, with TR1/TR2/TE = 70/120/2.5 ms, FOV = 128 × 128 mm², matrix size = 64 × 64; in-plane resolution = 2 × 2 mm², slice thickness = 5 mm, and averages = 5. The 2D AFI implementation was corrected for the slice profile imperfection by using a scale factor of 1.5, as determined by a phantom calibration study. Flow compensation gradients were applied along both the slice and imaging readout directions to reduce flow artifacts during B₁+ shimming and B₁+ mapping acquisitions. An example of the B₁+ optimization ROI and associated flip angle map is shown in Figure 2.

Figure 2.

Relative B₁+ map and histogram of normalized B1+ obtained from within the kidney for a representative subject. The ROI for evaluating the B₁+ distribution is illustrated with an anatomic image (inset on the left image). C.V. represents coefficient of variance, calculated as the ratio of standard deviation over mean.

Image Processing and Data Analysis

Voxel-wise T₁ and T₂ mapping and region of interest (ROI)-based model fitting for T₁ and T₂ estimations were all performed in Matlab 7.1 (MathWorks, Natick, MA). The statistical analysis was performed in GraphPad Prism 5.0 (GraphPad Software Inc., La Jolla, CA). ROIs for renal cortex and medulla were conservatively defined to reduce partial volume effects. For T₁ measurements, ROIs were determined by using images acquired with an inversion time (TI) near the nulling point for the medulla. For T₂ measurements, ROIs were defined on the T₂-weighted images that had the greatest contrast between the cortex and medulla. To further minimize partial volume effects and errors due to sporadic physiological noise, a trimmed mean was used for T₁ and T₂ estimations, excluding the 5% of voxels with the lowest values and the 5% with the highest values based on the histogram from each ROI (44). The number of voxels included in the renal cortex and medulla ROIs combined for a single kidney were 98 ± 30 and 118 ± 32 for 3T and 7T, respectively. For the estimation of background noise, 50 voxels were included.

T₁ and T₂ Mapping and Estimation

T₁ and T₂ fitting were performed on a pixel-wise basis to produce relaxometric parametric maps. To reduce the bias for T₁ and T₂ estimations due to the relatively low SNR near the nulling point for the IR preparation and at long effective TEs, data were fit using models that incorporated image noise, as described previously (35). For images acquired at different inversion times, nonlinear least square model fitting was performed with the following equation:

$$S(TI)=\sqrt{[S_0{(1-2{exp}^{(-TI/T_1)}+{exp}^{(-T_r/T_1)}]}^2+C_{\mathit{noise}}^2},$$ [1]

where S(TI) is the measured signal intensity at inversion time TI, S₀ the proton density image weighted by coil sensitivity and imaging gain, T₁ the tissue longitudinal relaxation time to be estimated, T_r the recovery time after imaging saturation and C_noise the estimated imaging noise. For images acquired with different TEs, the following equation was used:

$$S(TE)=\sqrt{{[S_0{exp}^{(-TE/T_2)}]}^2+C_{\mathit{noise}}^2},$$ [2]

where S(TE) is the measured signal intensity at each TE and T₂ is the tissue transverse relaxation time constant to be estimated.

Statistical Analysis

Repeated measures two-way analysis of variance (ANOVA) was performed to determine statistical significance of differences in relaxometry values between renal tissue types and between field strengths. The Bonferroni post-test following ANOVA was applied for multiple comparisons between renal tissue types and field strengths. Statistical significance was defined as a P-value less than 0.05.

The reproducibility of T₁ and T₂ values measured at 7T was evaluated as percent difference between two sessions by using the following formula:

$$\mathit{Difference}(\%)=\frac{(S_1-S_2)}{(S_1+S_2)/2}\times 100,$$ [3]

where S₁ and S₂ are renal T₁ and T₂ measurements from the first and second sessions.

RESULTS

T₁ and T₂ relaxometry measurements were successfully performed in all 7 subjects. Examples of 7T renal T₁ and T₂ mapping from a representative subject are presented in Figure 3. The estimated individual renal T₁ and T₂ values from 3T and the first session of 7T studies are presented in Figures 4a–4b, and group means and standard deviations of renal T₁ and T₂ values are listed in Table 1.

Figure 3.

7T T₁ and T₂ imaging results from one typical subject: (a) ss-FSE images were acquired by using six inversion times 100, 150 300, 500, 800 and 1200 ms; (b) ss-FSE images were acquired by using six different effective echo times: 20, 40, 80, 120, 160, and 200 ms. The conservatively defined ROIs (blue for the cortex and red for the medulla) used for the estimation of renal T₁ and T₂ are overlaid on the T₁ and T₂ maps generated from pixel-wise fitting.

Figure 4.

Renal T₁ and T₂ measurements at 3T and 7T (a–b) and the reproducibility of renal T₁ and T₂ measurements at 7T (c–d) from five subjects. The presented renal T₁ and T₂ measurements at 7T in (a–b) are from the first session studies.

Table 1.

Renal T₁ and T₂ measurements at 3T and 7T (mean ± S.D. ms) and percentage differences between two sessions at 7T from five subjects.

3T		Cortex		Medulla
		T1	T2	T1	T2
		1261 ± 86	121 ± 5	1676 ± 94	138 ± 7
7T	1st Session	1661 ± 68	108 ± 7	2094 ± 67	126 ± 6
	2nd Session	1632 ± 49	106 ± 9	2091 ± 51	122 ± 6
	Difference (%)	1.8 ± 4.9	1.7 ± 4.7	0.1 ± 3.7	3.7 ± 4.5

Repeated measures two-way ANOVA analysis showed no interaction effects between renal tissue type and field strength (P values = 0.83 and 0.73 for renal T₁ and T₂) but significant effects of these two factors on renal T₁ and T₂ values (P values <0.001 for both). Bonferroni post-test analysis showed significant differences between 3T and 7T for cortical and medullary T₁ (P values < 0.001) and T₂ (P values < 0.01 for renal cortex and < 0.05 for renal medulla), and between renal cortex and medulla for T₁ and T₂ measurements on 3T (P values < 0.001 for renal T₁ and < 0.01 for renal T₂) and 7T (P values < 0.001 for both renal T₁ and T₂).

Two-tailed paired t-tests showed insignificant differences in measured T₁ between the two 7T sessions (P values = 0.47 and = 0.94 for cortical and medullary T₁) and T₂ values (P values = 0.51 and = 0.15 cortical and medullary T₂). The 7T measurements of T₁ and T₂ were very reproducible with all group mean percent differences within 4% (Table 1 and Figure 4c–4d).

DISCUSSION

This is the first study reporting normal renal T₁ and T₂ values at 7 T. While similar methods have been used in renal relaxometry studies at 3T, a number of adaptations were needed to overcome the challenges of ultra-high field MR, including increased B₁ and B₀ inhomogeneity, short-term SAR constraints, and limited RF peak power.

Having a uniform and efficient B₁+ shim was required in order to produce a uniform inversion for the T₁ measurements. With the B₁+ shimming and calibration techniques used here, the B₁+ variation over the kidney was small, with a typical coefficient of variation of ~8% (Figure 2). We verified that this was sufficient to provide uniform inversion with an adiabatic inversion pulse by observing that 1) scans acquired with the higher HS4 RF voltage did not significantly change renal T₁ measurements and 2) no significant differences were found between T₁ measurements in regions with high and low B₁+ amplitudes (data not shown). The B₁+ optimization was made substantially easier by focusing on a signal kidney with axial imaging slices.

Although using single breath-hold ss-FSE helps reduce respiration motion artifacts and is insensitive to B₀ inhomogeneity, short-term SAR is a limiting factor, especially as the ss-FSE readout is combined with the additional inversion and refocusing pulses used for T₁/T₂ encoding. A combination of multiple RF power-reducing strategies was used to minimize short-term SAR. First, the number of phase encoding lines was minimized by using partial Fourier imaging as well as a high parallel imaging acceleration factor of 4. Second, the hyperecho technique was used to reduce the flip angle of the readout pulses (41). Third, the VERSE technique was used to reduce the peak power requirements for both excitation and refocusing pulses. Fourth, the T₁ measurement used interleaved long and short inversion times to better distribute the power deposition over time; likewise the T₂ measurement interleaved large and small numbers of preparation echoes. Finally, the parallel imaging reference scans were acquired only once per breath-hold rather than with each encoded acquisition. Since the required power varied between subjects, the shortest repetition time allowable under the short-term SAR limits was used in order the minimize breath-hold times.

To avoid partial volume effects on T₁ and T₂ estimations, conservative regions in the renal cortex and medulla were selected for data analysis. Furthermore, trimmed mean values within ROIs were used for final estimation to further remove possible errors due to partial volume effects and/or physiological noise.

Because of the challenges of performing relaxometry studies in the kidney, reported renal T₁ and T₂ in human are quite limited. Estimated 3T cortical and medullary T₁ values of the present study are comparable to those reported in the literature (1142 ± 154 ms and 1545 ± 142 ms for the cortex and medulla, respectively), but T₂ values are significantly higher than those previously reported (76 ± 7 ms and 81 ± 8 ms for the cortex and medulla, respectively) (35).

In the current study, T₂ contrast was achieved by using a Carr-Purcell-Meiboom-Gill (CPMG) style refocusing echo train rather than a single spin echo preparation (35). It is well known that the continuous application of refocusing RF pulses reduces signal loss due to molecular diffusion, field inhomogeneity and J coupling (45), resulting in longer T₂ estimates. T₂ values were indeed shorter at 3T when using a single spin echo preparation in both phantom and in vivo studies (data not shown). Estimates of renal T₂ values obtained by FSE imaging methods at 1.5T in another study (46) with reported mean T₂ values of 112 and 143 ms for the renal cortex and medulla, were also higher than the results using a single spin echo T₂ encoding (35). Since in practice, FSE imaging methods, such as segmented FSE and ss-FSE, are widely used for renal MRI, the protocol optimization for such imaging methods should be based on T₂ measurements that are estimated under similar conditions, which was the reason that the current strategy was used for estimating T₂.

It has been demonstrated that renal T₁ and T₂ imaging can be successfully performed at 7T in a single breath-hold, and that these T₁ and T₂ measurements were highly reproducible. However, due to short-term SAR constraints, mapping could only be achieved in a single kidney. To perform bilateral kidney relaxometry, multiple breath holds can be performed, but may be challenging for subjects in a clinical setting. Therefore, imaging methods with low RF power deposition suitable for T₁ and T₂ imaging are needed for UHF applications. It should also be noted, however, that even at lower field strengths, T₁ and T₂ mapping of the kidneys has been limited to a single slice acquisitions (29,30,46). The ability to obtain full coverage of the kidneys is still limited by respiratory motion and compounded by the challenge of accurate co-registration of images with varying contrast.

CONCLUSION

In summary, renal longitudinal and transverse relaxation times can be reliably and reproducibly measured by using an ss-FSE measurement strategy within a single breath-hold at 7T.

Abbreviations used

ANOVA

ANalysis Of VAriance

ASL

Arterial Spin Labeling

AFI

Actual Flip-angle Imaging

CPMG

Carr-Purcell-Meiboom-Gill

FA

Flip Angle

FOV

Field Of View

GRAPPA

GeneRalized Autocalibrating Partially Parallel Acquisitions

MRI

Magnetic Resonance Imaging

ROI

Region Of Interest

ss-FSE

single shot fast spin echo

S.D

Standard Deviation

SAR

Specific Absorption Rate

SNR

Signal Noise Ratio

TE

Echo Time

TR

Repetition Time

TrueFISP

TRue Fast Imaging with Steady Precession

UHF

Ultra High Field

VERSE

VariablE-Rate Selective Excitation