Pulse Sequence for Multislice T 1 -Weighted MRI

Transcription

1 Magnetic Resonance in Medicine 51: (2004) Pulse Sequence for Multislice T 1 -Weighted MRI Andrew J. Wheaton,* Arijitt Borthakur, Sridhar R. Charagundla, and Ravinder Reddy A 2D multislice spin-lock (MS-SL) MR pulse sequence is presented for rapid volumetric T 1 -weighted imaging. Image quality is compared with T 1 -weighted data collected using a singleslice (SS) SL sequence and T 2 -weighted data from a standard MS spin-echo (SE) sequence. Saturation of longitudinal magnetization by the application of nonselective SL pulses is experimentally measured and theoretically modeled as T 2 decay. The saturation data is used to correct the image data as a function of the SL pulse duration to make quantitative measurements of T 1. Measurements of T 1 using the saturation-corrected MS-SL data are nearly identical to those measured using an SS-SL sequence. The MS-SL sequence produces quantitative T 1 maps of an entire sample volume with the high-snr advantages conferred by SE-based sequences. Magn Reson Med 51: , Wiley-Liss, Inc. Key words: MRI; T 1 ; spin-lock; multislice; saturation In this study, spin-lock (SL) MRI was used to generate an alternative contrast to conventional proton density, T 1 -, or T 2 -weighted MRI methods. In SL-MRI, a long-duration, low-power SL pulse is applied to lock the spins in the transverse plane. During the SL pulse, the transverse magnetization decays according to T 1, the spin-lattice relaxation in the rotating frame of reference. The amplitude of the SL pulse is commonly referenced in terms of the nutation frequency ( B 1 ), which is typically in the range of a few hundred hertz to a few kilohertz. T 1 relaxation phenomena are sensitive to physicochemical processes with inverse correlation times on the order of the nutation frequency of the SL pulse. By setting the amplitude of the SL pulse to coincide with the frequency of the molecular processes of interest, the signal from the SL-MRI sequence becomes heavily weighted by the T 1 parameter according to Eq. [1], where TSL is the SL pulse duration, and S is the signal intensity as a function of TSL. S TSL S 0 e TSL/T1 [1] By acquiring a series of T 1 -weighted images at varying TSL durations with constant SL pulse amplitude, and using Eq. [1], the T 1 parameter can be measured on a pixelby-pixel basis using linear regression to create a quantitative spatial map of T 1 values. SL pulse sequences may be applicable for imaging multiple systems, such as muscle (1,2), breast (3), liver (4), Department of Radiology, University of Pennsylvania, Philadelphia, Pennsylvania. Grant sponsor: NIH; Grant number: RR02305; Grant sponsor: NIAMS-NIH; Grant number: R01-AR45404; Grant sponsors: Arthritis Foundation; Whitaker Foundation. *Correspondence: Andrew J. Wheaton, B.S., Department of Radiology, University of Pennsylvania, B1 Stellar-Chance Laboratories, 422 Curie Blvd., Philadelphia, PA wheaton@seas.upenn.edu Received 19 August 2003; revised 30 September 2003; accepted 3 October DOI /mrm Published online in Wiley InterScience ( Wiley-Liss, Inc. 362 brain (5,6), and tumors (7,8). T 1 relaxation has been used as an indicator of proteoglycan content in articular cartilage, and is being developed as a diagnostic tool for the detection of osteoarthritis (9 12). In addition, T 1 - weighted MRI of H 2 17 O was recently used to measure cerebral and tumor perfusion (13,14). The sensitivity of T 1 to a variety of physiological parameters has also been demonstrated (15). T 1 weighting can be added to most MRI sequences by including an SL pulse cluster at the beginning of the pulse sequence. The magnetization thereby becomes T 1 -prepared and results in a T 1 -weighted signal. A 2D singleslice (SS) SL sequence based on a spin-echo (SE) sequence has been successfully implemented to produce a single 2D T 1 -weighted slice (16). The schematic for the SS-SL sequence is shown in Fig. 1. The nutation of the magnetization in this scheme is summarized in the rotating frame vector diagrams in Fig. 2. A single slice of the T 1 -prepared magnetization is excited using a 90 slice-selective pulse followed by a conventional SE readout. More recently, a T 1 -prepared 3D SL gradient-echo (GE) sequence was developed (17). In the 3D sequence, the magnetization is T 1 -prepared in the same way as the SS-SL sequence. However, in place of the 90 slice-selective pulse, a nonselective low-angle pulse is used to excite a slab, followed by a conventional 3D GE readout. Like most 3D GE-based sequences, the 3D T 1 sequence produces images with a reduced signal-to-noise ratio (SNR) compared to SE-based 2D images. Furthermore, in the interest of time, a short TR is typically employed, which introduces significant T 1 weighting in the resultant image and obscures T 1 -based contrast. Given the desirability of volumetric T 1 imaging, an MS-SL sequence would be useful. With an MS sequence, multiple 2D T 1 -weighted slices can be collected and the acquisition times are similar to those of SS sequences. With an MS-SE-based sequence, an entire object of interest can be imaged with higher SNR than a comparable 3D GE-based sequence. The nonselective nature of the hard pulses in T 1 preparation used in the current implementation of the SS-SL sequence prevents its use in multislice imaging because the non-imaged slices are significantly saturated, which leads to severe artifacts. A novel method for slice-selective spin-locking is required to prevent excessive saturation of magnetization during an MS-SL acquisition. The saturation of longitudinal magnetization from repeated RF pulses in standard multislice sequences has been previously observed (18,19). In the MS-SL sequence, the saturation effect is expected to be more pronounced due to the inclusion of long-duration SL pulses. Recently, a phase-alternating, or self-compensating, SL pulse was developed to reduce signal decay of nonresonant spins (20). The concept is based on a rotary echo (21) consisting of two halves with equal amplitude and opposite phase

2 Multislice T 1 MRI 363 theoretically. Saturation-compensated MS-SL images are used to compute T 1 maps that are nearly identical to those obtained with a validated SS-SL imaging sequence. MATERIALS AND METHODS MS-SL Pulse Sequence Design FIG. 1. 2D SS-SL sequence with SE readout. The RF pulses and gradients are only performed once per TR. ( 90 ). The self-compensating pulse is theorized to restore magnetization that is not in the transverse plane to its original orientation after the initial excitation (20). Its effect is chiefly due to the restoration of incompletely spin-locked spins resulting from an inhomogeneous B 1 field. The single-phase SL pulse is unable to refocus the improperly spin-locked magnetization, and therefore it produces a degraded signal. The signal preservation characteristics of the self-compensating SL pulse are conducive toward its application in the MS-SL sequence. A comparison of the effects of a single-phase SL pulse vs. a self-compensating SL pulse is shown in Fig. 3. The data from the single-phase SL pulse exhibit rapid decay in comparison to the self-compensating SL pulse. Despite the use of a self-compensating SL pulse, there still exists residual signal loss attributed to saturation of the longitudinal magnetization by the SL pulse. Therefore, images produced with the MS-SL sequence will have a combination of T 1 weighting and saturation weighting. This result is acceptable if the primary purpose is to create multislice images with T 1 contrast, and the saturation effect is small. However, to make a quantitative measurement of T 1, saturation decay must be accounted for. In this work we describe a new pulse sequence for multislice T 1 -weighted imaging. Signal decay of nonimaged slices is reduced by using a self-compensating SL pulse. Saturation of the longitudinal magnetization by the SL pulses is also measured experimentally and modeled The MS-SL pulse sequence was based on a conventional 2D MS-SE design. A schematic for the pulse sequence is shown in Fig. 4. The magnetization from the longitudinal axis was flipped into the transverse plane using a sliceselective 90 pulse. The magnetization was T 1 -prepared using an on-resonance, nonselective, low-power, long-duration, self-compensating SL pulse. The magnetization was then phase-encoded and read using a standard 2D SE acquisition scheme. The echo time (TE) was calculated in reference to the end of the SL pulse (i.e., a delay equal to half of the TE was followed by a slice-selective 180 pulse). After another delay of half of the TE, the signal was acquired. The pulse sequence was repeated N times within a given TR to acquire N slices. We chose to implement an SE version of the MS-SL sequence; however, a GE-based readout can also be employed. Saturation Model We developed a theoretical model to describe the saturation effect of repeated nonselective SL pulses on longitudinal magnetization. We focused our analysis on the longitudinal magnetization because the application of an SL pulse on transverse magnetization will produce the wellcharacterized result of T 1 decay. Our model is based on the general analytical solution to the Bloch equations as described by Madhu and Kumar (22). In brief, the model solves the Bloch equations in the rotating frame as a function of time for the given input parameters of 1, T 1, T 2, and ; 0. In particular, we examined the special case of the effect of an on-resonance ( 0 0) RF pulse on longitudinal magnetization. For an on-resonance RF field with the given input parameters, the saturated magnetization, M sat, is described by Eq. [2] with the equilibrium term sat M equilibrium, coefficients c 1,c 2,c 3, and with constant terms 1, 2, 3. M sat sat t M equilibrium c 1 e 1t e 2t c 2 cos 3 t c 3 sin 3 t [2] FIG. 2. Vector diagram of the nutation of the magnetization in the rotating frame of reference. The magnetization (M) begins along the z-axis in the equilibrium position (a) and is nutated into the y-axis by the initial hard 90 along the x-axis. The magnetization then nutates in the rotating frame about the SL pulse applied along the y-axis (b). At the end of the SL pulse duration (TSL), the magnetization is left in the transverse plane (c). The magnetization is then restored to the longitudinal axis by the second hard 90 applied along the negative x-axis (d). The resultant T 1 -prepared magnetization is ready to be read using a GE or SE readout. Residual transverse magnetization is destroyed using a crusher gradient immediately following the second hard 90 pulse.

3 364 Wheaton et al. at the end of the pulse is decayed by the cosine modulation. Therefore, the saturation of the magnetization by an on-resonance, self-compensating SL pulse can be most simply described by T 2 decay as a function of TSL (Eq. [6]). M sat TSL e TSL/T2 [6] During the period between each SL pulse TR/N, the longitudinal magnetization recovers according to T 1. After each delay, the magnetization becomes saturated by a factor of M sat. Therefore, at the end of the Nth regrowth delay period, the longitudinal magnetization, M z,n, is given by the recursive Eq. [7], where M sat is determined according to Eq. [6]. M z,n M sat M z,n 1 e /T1 M 0 1 e /T1 [7] FIG. 3. Comparison of saturation effects on longitudinal magnetization using a self-compensating SL pulse and a single-phase SL pulse as a function of TSL. Signal intensities are normalized in each case with respect to unsaturated data (TSL 0). Saturation occurs much more rapidly with a single-phase SL pulse. The coefficients and constant terms are cumbersome expressions that are detailed in Ref. 22 and hence are not reproduced here. The equilibrium term is identical to the steady-state equation for on-resonance saturation (Eq. [3]) (23). sat M equilibrium 1 1 B 1 2 [3] T 1 T 2 In the limiting case of on-resonance saturation with T 1 on the order of seconds, T 2 on the order of milliseconds, sat and B 1 on the order of kilohertz, c 1,c 3, and M equilibrium approach zero, c 2 approaches 1, 3 approximates 1, and 2 reduces to 1/T 2 which is the reciprocal average of T 1 and T 2 (Eq. [4]) (21), thereby yielding a simplified result as a function of TSL (Eq. [5]). Using Eq. [7] in conjunction with the known or measured B 1,, TSL, T 1, and T 2 parameters, the longitudinal magnetization at the end of the train of N-1 SL pulses and N- delays can be predicted. Spectroscopy Experiments All MR data were acquired on a Varian INOVA 4.7 T research console using a 2-cm-diameter cylindrical birdcage coil. The sample was a cylindrical homogeneous agarose phantom (4% weight/volume). T 1 and T 2 of the phantom were measured to be 880 ms and 33 ms, respectively, by means of inversion recovery and progressive saturation experiments. A global measurement of T 1 was performed using a T 1 spectroscopy sequence. Spectra were acquired for 11 TSL times evenly spaced from 0 to 100 ms. The magnetization was allowed to fully relax between each acquisition. The nutation frequency of the SL pulse was set to 500 Hz. Data were measured as the intensity of the peak of each spectrum. The data were normalized with respect to the intensity of the spectrum acquired with zero SL pulse duration. T 1 was measured as the linear regression of the normalized data based on Eq. [1]. To measure 1 T T 1 1 T 2 [4] M sat TSL cos 1 TSL e TSL/T2 [5] The simplified model is in agreement with the analytical solution describing the effect of an on-resonance SL pulse on longitudinal magnetization in Ref. 20. The cosine term in Eq. [5] represents the modulation of the non-spin-locked magnetization about the B 1 axis that results in image artifacts. It is from these equations that the importance of using a self-compensating SL pulse can be discerned mathematically. During the first half of the selfcompensating pulse, the magnetization will rotate by 1 TSL/2. During the second half of the self-compensating SL pulse, the 1 TSL/2 rotation is refocused and the angle dependence is subsequently eliminated as the cosine term reduces to unity (20). The single-phase SL pulse will not refocus the initial 1 TSL/2 rotation, and hence the signal FIG. 4. 2D MS-SL sequence with SE readout. The events within the brackets are repeated N times within a given TR for the acquisition of N slices.

4 Multislice T 1 MRI 365 were compared using two-tailed Student s t-tests at a 5% confidence level, using the JMPIN statistical package. FIG. 5. Spectroscopic pulse sequence to measure the saturation effect of the MS-SL sequence. To measure the saturation from an MS-SL acquisition of N slices within a given TR for a given TSL, a train of N-1 self-compensating SL pulses separated by delays of TR/N precede an SE readout. The pulse sequence for N 3is shown. the saturation effect of the SL pulse on longitudinal magnetization, a simple pulse sequence consisting of an SL pulse immediately followed by an SE readout was used. The acquisition parameters and data processing were identical to those in the T 1 spectroscopy experiment. We experimentally measured the saturation effect of the train of nonselective SL pulses in the MS-SL sequence using a spectroscopy pulse sequence (Fig. 5). The pulse sequence consisted of a train of N-1 self-compensating SL pulses separated by delays of TR/N, followed by a nonselective SE readout. The sequence was run twice for N 3 and N 5 (both with TR 2000 ms). Spectra were acquired for 11 TSL times evenly spaced from 0 (unsaturated) to 100 ms (most saturated). The longitudinal magnetization was allowed to fully recover between each acquisition. Data were normalized with respect to the TSL 0 data point in the same manner as before. Experimental data were correlated with the theoretical predictions from the model using linear regression (JMPIN statistical package; SAS Institute, Cary, NC). Phantom Image Acquisition The MS-SL imaging sequence was employed with the following parameters: TE/TR 13/2000 ms, FOV 15 mm 15 mm, acquisition matrix , slice thickness 2 mm, and two signal averages, for a total acquisition time of 8 min 50 s per image set. A self-compensating SL pulse was used with a nutation frequency of 500 Hz. A series of images were acquired for four TSL times evenly spaced from 10 ms to 40 ms. The acquisition of a complete TSL series was repeated for image sets of one, three, and five slices. To illustrate the necessity of using the self-compensating SL pulse, a multislice SL sequence pre-encoded with a single-phase SL pulse was used to collect a fiveslice image set using identical imaging parameters. An SE image set of five slices without T 1 preparation was acquired using a standard MS-SE sequence with TE 23 ms. In addition, a single slice in the center of the object was acquired using identical image settings and SL pulse parameters employing the SS-SL sequence detailed in Fig. 1. Images were reconstructed and processed using a custombuilt visualization package written in IDL (RSI, Boulder, CO). Assuming homogeneity of the phantom, image signal intensities were measured as the average of a region of interest (ROI) drawn over the entire phantom. The ROIs were identical for all images. Average signal intensities Saturation Correction and T 1 Measurement The raw phantom images were corrected for saturation as a function of TSL and N using the data collected from the saturation experiments. Each image in the TSL series was divided by the normalized saturation value to create a saturation-compensated image. The corrected images were fit as a function of TSL on a pixel-by-pixel basis using linear regression according to Eq. [1] to create a spatial map of T 1 values. For comparison, the raw multislice SL images were also corrected using the saturation data predicted by the model. The image series were likewise fit using linear regression to measure T 1. The true value of T 1 was measured from the SS-SL data. The use of the SS-SL sequence to measure the true T 1 was validated by the spectroscopy measurement. Assuming homogeneity of the relaxation properties of the agarose phantom, T 1 was measured as the average of an ROI drawn over the entire phantom. ROIs were identical for all images. Average T 1 values were compared using two-tailed Student s t-tests at a 5% confidence level, using the JMPIN statistical package. In Vivo Image Acquisition To illustrate the application of the MS-SL sequence in vivo, a T 1 -weighted image set was acquired in the brain of a healthy mouse. The mouse was handled according to the guidelines of the Institutional Animal Care and Use Committee at the University of Pennsylvania. The MS-SL imaging sequence was employed using parameters of TE/ TR 13/2000 ms, TSL 10 ms, B Hz, FOV 30 mm 30 mm, acquisition matrix , five slices, slice thickness 1 mm, and two signal averages, for a total acquisition time of 8 min 50 s. For comparison, a single image of the mouse brain was obtained using the FIG. 6. Plot of saturation as a function of SL pulse duration (TSL) for both experimentally collected data and model predictions. There was excellent agreement between the theoretical model and experimental data (slope 0.97, R , P 0.001).

5 366 Wheaton et al. FIG. 7. Complete resultant image sets from (a) SS-SL, (b) MS-SL (N 1), (c) MS-SL (N 3), (d) MS-SL (N 5), (e) MS-SE sequence (N 5), and (f) MS-SL collected using a single-phase SL pulse (N 5). All images are identically scaled. SS-SL sequence with imaging parameters identical to those used in the MS-SL acquisition. RESULTS Saturation Data and Model Spectral data collected using self-compensating SL pulses are shown in Fig. 6 as a function of TSL. The model predictions are overlaid on the experimental data for comparison. The model was strongly correlated with the experimental data (slope 0.97, R , P 0.001). The exponential decay of the data as a function of TSL is caused by saturation from the SL pulse. The greater the duration of the SL pulse, the greater the saturation of the longitudinal magnetization. The decay constant of the curve matches T 2 according to the model. Phantom Image Data For all cases, images collected using the MS-SL sequence were void of artifacts (Fig. 7). The image quality of the MS-SL images was qualitatively similar to that of SS-SL and MS-SE-generated images. Images collected using the single-phase SL pulse exhibited considerable artifacts and signal loss compared to the self-compensating SL images. A comparison of the average signal intensity from each FIG. 8. Comparison of raw signal intensities for SS-SL, MS-SL (N 1), MS-SL (N 3), MS-SL (N 5), MS-SE (N 5) (MS-SE 5), and MS-SL with single-phase SL pulse (N 5) (MS-SL single-phase SL). Error bars represent the standard deviation of the signal intensity within the ROI drawn on the homogeneous phantom. Note the dramatic improvement of the MS-SL image intensity in comparison with the MS-SL single-phase SL data. FIG. 9. Normalized average image signal intensity as a function of TSL for SS-SL, MS-SL without saturation correction, and MS-SL with saturation correction based on model predictions. The case of N 5 is shown, but the curves are representative of the N 3 case.

6 Multislice T 1 MRI 367 Table 1 Measurements of T 1 for SS SL Sequence and MS SL Sequence for 1, 3, and 5 slices, Respectively* T 1 Measurements A B C SS SL 45.7 ms N/A N/A 2D MS SL ms N/A N/A 2D MS SL ms 45.2 ms 43.3 ms 2D MS SL ms 44.8 ms 43.7 ms *A, estimate of T 1 without using saturation correction; B, estimate of T 1 using saturation correction based on experimental data; C, estimate of Ts 1 using saturation correction based on model prediction. The corrected estimates are not applicable for SS SL and MS SL (N 1) because saturation correction was not applicable to these data. data set acquired with the same effective TEs (TSL TE 10 ms 13 ms 23 ms) is displayed in Fig. 8. All images had approximately the same noise factor, and thus SNR is proportionate to signal intensity. The signal intensities of the SS-SL, and the one- and three-slice MS-SL sequences are statistically indistinguishable according to a two-tailed Student s t-test (P 0.25 and P 0.12, respectively). Signal intensity decreases with increasing number of slices in the MS-SL sequence due to the increase of the saturation effect with increasing number of slices. The average signal intensity of the five-slice MS-SL sequence is statistically different from that of the five-slice MS-SE sequence (two-tailed Student s t-test; P 0.01). The signal intensities of the SL images were less than those of the SE images due to the saturation effect from the five SL pulses within each TR. Measurement of T 1 Normalized average signal intensities with and without saturation correction as a function of TSL for the five-slice MS-SL data are shown in Fig. 9. The SS decay curve was strongly correlated with the saturation model-corrected data (R , P 0.001). The average T 1 values for all of the data are summarized in Table 1. The SS-SL data yields the true measurement of T 1, as validated by the spectroscopic measurement. Without saturation compensation, the uncorrected signal intensities decay more quickly due to the increasing saturation effect with increasing TSL. Hence, the estimation of T 1 using the uncorrected data was considerably lower than the true measurement (Table 1). However, with saturation compensation, the average T 1 measurement was nearly the same as the true measurement of 45.7 ms (two-tailed Student s t-test; P 0.68 for MS-SL 3, and P 0.47 for MS-SL 5). T 1 measurements using the model-generated data to account for saturation were also similar to both the true measurement (P 0.06 for MS-SL 3, and P 0.11 for MS-SL 5) and the measurement using saturation correction based on experimental data (P 0.12 for MS-SL 3, and P 0.37 for MS-SL 5). In Vivo Image Data The in vivo image set of the mouse brain collected using the MS-SL sequence was void of artifacts and variation in signal intensity from slice to slice (Fig. 10). The image data from the MS-SL sequence were qualitatively similar to the image data collected with the SS-SL sequence (Fig. 11). The signal intensity in the MS-SL image is reduced in comparison to the SS-SL image due to saturation from the repeated application of SL pulses in the MS-SL sequence. The average signal intensity of the brain in the MS-SL image is 30% less than in the SS-SL image. For comparison, the average signal intensity of the muscle tissue was reduced by 50%. The increased signal reduction in the muscle tissue in comparison to the brain is due to differences in T 2 decay. The SS-SL and MS-SL images have identical T 2 and T 1 weighting, whereas the MS-SL image has an additional decay factor of T 2 due to the saturation effect. Since muscle tissue has a shorter T 1 and T 2 than brain, it also has a shorter T 2, which results in the increased signal decay evident in the MS-SL images. DISCUSSION The MS-SL sequence produced images that were statistically similar in signal intensity in comparison with images collected using SS-SL and MS-SE pulse sequences. For reasonable lengths of TSL, the MS-SL sequence can be used to create T 1 -weighted images with good SNR. The pre-encoding scheme is versatile enough to be attached to most common MR pulse sequences. A limitation of the MS-SL sequence is the signal decay from persistent saturation at long TSL. The use of a self-compensating SL pulse mitigates signal loss due to saturation decay and B 1 inhomogeneity, which severely corrupt images, as evident in images acquired with a single-phase SL sequence. Although a single-phase SL pulse has been successfully implemented in previous versions of the SS T 1 MRI sequence, in the MS-SL sequence, the repeated application of the SL pulse within each TR necessitates the use of a self-compensating SL pulse. FIG. 10. MS-SL images of in vivo mouse brain. For convenience, three images are shown from the five-slice image set.

7 368 Wheaton et al. FIG. 11. a: SS-SL image of an in vivo mouse brain. b: Corresponding image from the five-slice MS-SL image set collected using identical imaging parameters. Both images are scaled identically. The brain is identified by an arrow in both images. Muscle tissue is encompassed by a solid ROI in both images. The raw MS-SL images contain a combination of T 1 weighting and saturation weighting. If the goal is simply to produce an image with T 1 contrast, the MS-SL sequence can be used to create images with T 1 weighting without regard to saturation. However, in order to make an accurate measurements of T 1, the saturation effect must be taken into account. By compensating for this effect in the images using the saturation data, the T 1 parameter can be quantitatively calculated. The close agreement of the estimates of T 1 from the MS-SL data and the true measurement of T 1 validate the MS-SL sequence with saturation compensation as an accurate method to quantitatively measure T 1. The saturation model was strongly correlated with the experimental data (R , P 0.001). The strength of the model suggests that if T 1 and T 2 are known for the sample, the saturation data can be accurately predicted. The saturation data can be used to correct an image series in order to make a quantitative measurement of T 1. With a general analytical model, experimentally obtained saturation data are not necessary to make an accurate measurement of T 1. In this work, based on the assumption of a homogeneous sample, the saturation data were measured globally. Although possible local variation in T 2 was not taken into account, the measurements of T 1 using the globally-corrected saturation data coincided with the spectroscopy-validated measurement. However, for application in biological tissues with spatially inhomogeneous relaxation properties, such as the mouse brain reported in this work, T 2 must be spatially resolved in order to perform saturation correction on a pixel-by-pixel basis. Our group is currently investigating T 2 mapping techniques to collect saturation data to be used in conjunction with MS-SL data to create T 1 maps in biological tissue. The development of multislice SL imaging allows for the acquisition of a T 1 -weighted volume of interest without the limitations of the GE-based SL sequence (e.g., T 1 -dominated contrast and lower SNR). Multislice T 1 -weighted acquisitions would be particularly desirable for imaging the brain in studies of Alzheimer s disease. A recent study used T 1 MRI to detect amyloid- plaques, which are characteristic of Alzheimer s disease, in the brain of a mouse model (24). The ability of the MS-SL sequence to cover the brain in a single acquisition, along with the advantages of high SNR (which is necessary to detect small plaques), makes this sequence a viable option this application. CONCLUSIONS We developed a multislice SL imaging sequence for T 1 mapping, and validated its performance by experiments on a phantom. The MS-SL pulse sequence generates T 1 - weighted images with SNR and image quality similar to those in SE images. The saturation effect of the MS-SL sequence was experimentally measured and analytically modeled as T 2 decay. Saturation-corrected T 1 -weighted images were used to compute T 1. The theoretical model and experimental T 1 data are in excellent agreement. The MS-SL sequence can be used to measure T 1 in multiple 2D slices in about the same time as an SS acquisition. The SL scheme is versatile enough to be incorporated into many current pulse sequences. REFERENCES 1. Franczak MB, Ulmer JL, Jaradeh S, McDaniel JD, Mark LP, Prost RW. Spin-lock magnetic resonance imaging of muscle in patients with autosomal recessive limb girdle muscular dystrophy. J Neuroimaging 2000;10: Lamminen AE, Tanttu JI, Sepponen RE, Pihko H, Korhola OA. T1 rho dispersion imaging of diseased muscle tissue. Br J Radiol 1993;66: Santyr GE, Henkelman RM, Bronskill MJ. Spin locking for magnetic resonance imaging with application to human breast. Magn Reson Med 1989;12: Halavaara JT, Sepponen RE, Lamminen AE, Vehmas T, Bondestam S. Spin lock and magnetization transfer MR imaging of local liver lesions. Magn Reson Imaging 1998;16: Rizi RR, Charagundla SR, Song HK, Reddy R, Stolpen AH, Schnall MD, Leigh JS. Proton T1rho-dispersion imaging of rodent brain at 1.9 T. J Magn Reson Imaging 1998;8: Aronen HJ, Ramadan UA, Peltonen TK, Markkola AT, Tanttu JI, Jaaskelainen J, Hakkinen AM, Sepponen R. 3D spin-lock imaging of human gliomas. Magn Reson Imaging 1999;17: Poptani H, Duvvuri U, Miller CG, Mancuso A, Charagundla S, Fraser NW, Glickson JD, Leigh JS, Reddy R. T1rho imaging of murine brain tumors at 4 T. Acad Radiol 2001;8: Markkola AT, Aronen HJ, Paavonen T, Hopsu E, Sipila LM, Tanttu JI, Sepponen R. Spin lock and magnetization transfer imaging of head and neck tumors. Radiology 1996;200: Akella SV, Regatte RR, Gougoutas AJ, Borthakur A, Shapiro EM, Kneeland JB, Leigh JS, Reddy R. Proteoglycan-induced changes in T1rhorelaxation of articular cartilage at 4T. Magn Reson Med 2001;46: Mlynarik V, Trattnig S, Huber M, Zembsch A, Imhof H. The role of relaxation times in monitoring proteoglycan depletion in articular cartilage. J Magn Reson Imaging 1999;10: Duvvuri U, Reddy R, Patel SD, Kaufman JH, Kneeland JB, Leigh JS. T1rho-relaxation in articular cartilage: effects of enzymatic degradation. Magn Reson Med 1997;38:

8 Multislice T 1 MRI Regatte RR, Akella SV, Borthakur A, Reddy R. Proton spin-lock ratio imaging for quantitation of glycosaminoglycans in articular cartilage. J Magn Reson Imaging 2003;17: Tailor DR, Poptani H, Glickson JD, Leigh JS, Reddy R. High-resolution assessment of blood flow in murine RIF-1 tumors by monitoring uptake of H(2)(17)O with proton T(1rho)-weighted imaging. Magn Reson Med 2003;49: Tailor DR, Roy A, Regatte RR, Charagundla SR, McLaughlin AC, Leigh JS, Reddy R. Indirect 17O-magnetic resonance imaging of cerebral blood flow in the rat. Magn Reson Med 2003;49: Kettunen MI, Grohn OH, Silvennoinen MJ, Penttonen M, Kauppinen RA. Effects of intracellular ph, blood, and tissue oxygen tension on T1rho relaxation in rat brain. Magn Reson Med 2002;48: Duvvuri U, Charagundla SR, Kudchodkar SB, Kaufman JH, Kneeland JB, Rizi R, Leigh JS, Reddy R. Human knee: in vivo T 1 (rho)-weighted MR imaging at 1.5 T preliminary experience. Radiology 2001;220: Borthakur A, Wheaton A, Charagundla SR, Shapiro EM, Regatte RR, Akella SV, Kneeland JB, Reddy R. Three-dimensional T1rho-weighted MRI at 1.5 Tesla. J Magn Reson Imaging 2003;17: Melki PS, Mulkern RV. Magnetization transfer effects in multislice RARE sequences. Magn Reson Med 1992;24: Dixon WT, Engels H, Castillo M, Sardashti M. Incidental magnetization transfer contrast in standard multislice imaging. Magn Reson Imaging 1990;8: Charagundla SR, Borthakur A, Leigh JS, Reddy R. Artifacts in T(1rho)- weighted imaging: correction with a self-compensating spin-locking pulse. J Magn Reson 2003;162: Solomon I. Rotary spin echoes. Phys Rev Lett 1959;2: Madhu PK, Kumar A. Direct Cartesian-space solutions of generalized Bloch equations in the rotating frame. J Magn Reson Ser A 1995;114: Abragam A. Principles of nuclear magnetism. Oxford, UK: Oxford University Press; Borthakur A, Uryu K, Shively SB, Poptani H, Corbo M, Charagundla SR, Trojanowski JQ, Lee VM, Reddy R. In vivo T1 -weighted MRI of amyloid transgenic mouse model of Alzheimer s disease. In: Proceedings of the 11th Annual Meeting of ISMRM, Toronto, Canada, 2003.