Routine clinical brain MRI sequences for use at 3.0 Tesla

Authors

  • Hanzhang Lu PhD,

    1. Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    3. F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, Maryland, USA
    Search for more papers by this author
  • Lidia M. Nagae-Poetscher MD,

    1. Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, Maryland, USA
    Search for more papers by this author
  • Xavier Golay PhD,

    1. Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, Maryland, USA
    Search for more papers by this author
  • Doris Lin MD,

    1. Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Martin Pomper MD,

    1. Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Peter C.M. van Zijl PhD

    Corresponding author
    1. Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, Maryland, USA
    • Johns Hopkins University School of Medicine, Dept. of Radiology, 217 Traylor Bldg., 720 Rutland Ave., Baltimore, MD 21205
    Search for more papers by this author

Abstract

Purpose

To establish image parameters for some routine clinical brain MRI pulse sequences at 3.0 T with the goal of maintaining, as much as possible, the well-characterized 1.5-T image contrast characteristics for daily clinical diagnosis, while benefiting from the increased signal to noise at higher field.

Materials and Methods

A total of 10 healthy subjects were scanned on 1.5-T and 3.0-T systems for T1 and T2 relaxation time measurements of major gray and white matter structures. The relaxation times were subsequently used to determine 3.0-T acquisition parameters for spin-echo (SE), T1-weighted, fast spin echo (FSE) or turbo spin echo (TSE), T2-weighted, and fluid-attenuated inversion recovery (FLAIR) pulse sequences that give image characteristics comparable to 1.5 T, to facilitate routine clinical diagnostics. Application of the routine clinical sequences was performed in 10 subjects, five normal subjects and five patients with various pathologies.

Results

T1 and T2 relaxation times were, respectively, 14% to 30% longer and 12% to 19% shorter at 3.0 T when compared to the values at 1.5 T, depending on the region evaluated. When using appropriate parameters, routine clinical images acquired at 3.0 T showed similar image characteristics to those obtained at 1.5 T, but with higher signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), which can be used to reduce the number of averages and scan times. Recommended imaging parameters for these sequences are provided.

Conclusion

When parameters are adjusted for changes in relaxation rates, routine clinical scans at 3.0 T can provide similar image appearance as 1.5 T, but with superior image quality and/or increased speed. J. Magn. Reson. Imaging 2005;22:13–22. © 2005 Wiley-Liss, Inc.

IT IS BECOMING apparent that 3.0 Tesla (T) may turn into the new field strength of choice for clinical MRI. Advantages of higher magnetic field strength have already been demonstrated for many applications (1–9). An expected signal-to-noise ratio (SNR) increase proportional to the magnetic field strength is the most appealing feature of 3.0-T MRI (10), but other properties such as increased T1 relaxation time, decreased T2 relaxation time, increased magnetic susceptibility contrast, and increased spectral resolution for MR spectroscopy (MRS) may also provide great advantages. For example, increased SNR and longer T1 at 3.0 T allow better detection of aneurysms using three-dimensional time of flight (TOF) MR angiography (MRA), while increased SNR and larger susceptibility effects result in higher spatial resolution as well as lower contrast agent dosage required for contrast-enhanced MRA (1). Caveats include the fact that increased susceptibility can also increase image artifacts and distortions and that the radiofrequency (RF) power deposition is strongly increased due to its proportionality to the square of the magnetic field, which may compromise the use of certain pulse sequences.

Many of these features have been reported for specific applications. However, if 3.0 T is to become the clinical standard, it is important that the well-characterized image features relied upon for routine clinical diagnosis at 1.5 T can be consistently reproduced at 3.0 T. Thus, in view of the differences in T1 and T2 relaxation times between 1.5 and 3.0 T, it is necessary to adjust the image acquisition parameters for sequences providing optimal T1-weighted, T2-weighted and fluid-attenuated inversion recovery (FLAIR) contrast.

We determined T1 and T2 relaxation times (N = 10) for several major gray and white matter structures at 1.5 and 3.0 T, and used numerical simulation to calculate the required acquisition parameters for the routine clinical pulse sequences at 3.0 T. In the clinical setting, optimal contrast-to-noise ratio (CNR) is essential for diagnostic accuracy. However, CNR cannot be pursued at the expense of the study duration, since long scans are usually not well tolerated by patients. Accordingly, we have attempted to increase CNR using equal or shorter scan duration compared to 1.5 T. To test our protocols, we scanned 10 subjects, including five normal controls and five patients with various pathologies, and performed quantitative comparison for SNR and CNR.

MATERIALS AND METHODS

T1 and T2 Relaxation Time Measurements

A total of 10 normal subjects (four women, six men; age range 21–38 years, mean 28 ± 5 years) were scanned with both 1.5-T and 3.0-T MR systems (Intera; Philips Medical Systems, Best, The Netherlands). All subjects in this study were scanned after informed consent was obtained under protocols approved by the local Institutional Review Board.

A single oblique-axial slice (thickness = 6 mm) was positioned parallel to the anterior commissure–posterior commissure line at the level of the basal ganglia, and the location of the reference line was recorded to ensure slice position reproducibility for the two different field strengths. All the subjects were scanned by the same investigators (H.L., L.M.N.P.). Pulse sequences performed included an inversion recovery (IR) sequence (TI = 180, 630, 1170, 1830, 2610, 3450, 4320, 5220, 6120, and 7010 msec; recovery time [time between the excitation pulse and next inversion pulse] = 8000 msec; TE = 42 msec; field of view [FOV] = 220 × 176 mm; scan matrix size = 224 × 179; reconstructed matrix size = 256 × 256, number of signals averaged (NSA) = 1, gradient- and spin-echo [GRASE] acquisition (11) with echo-planar imaging [EPI] factor = 7; and turbo spin echo (TSE) factor = 4, scan duration = 13 minutes and 10 seconds) and a Carl-Purcell-Meiboom-Gill (CPMG) sequence (TR/TE = 1000/25, 50, 75, 100, 125, 150, 175, 200 msec; FOV = 220 × 176 mm; scan matrix size = 224 × 179; reconstructed matrix size = 256 × 256; NSA = 4; segmented EPI factor = 7; scan duration = one minute and 44 seconds]. The IR sequence was used to generate T1 maps and the CPMG sequence was used to generate T2 maps. The subjects were scanned by both systems on the same day or up to three days apart, following the same order for each subject, and semirandomly among the different subjects. The relaxation times were calculated on a voxel-by-voxel basis using in-house MATLAB (Mathworks, Natick, MA, USA) scripts. T1 and T2 relaxation times were measured by one investigator (L.M.N.P.) from the following regions of interest (ROI): genu and splenium of the corpus callosum, bilateral frontal and occipital white and gray matter, caudate, putamen, and thalamus (see Fig. 1 for examples). Cautions were taken to ensure that the ROIs on the 1.5-T and 3.0-T images were closely matched. The ROIs were drawn using the software ImageJ (National Institutes of Health, Bethesda, MD, USA).

Figure 1.

Absolute T1 (left column) and T2 (right column) maps acquired from a normal subject at 3.0 T (top row) and at 1.5 T (bottom row). Note that in the T2 map, the difference between gray matter and white matter is not as pronounced as in T2-weighted images, in which the combination of spin density effect and T2-weighting effect enhances the gray–white matter contrast. Location of ROIs is displayed on the bottom left. FGM = frontal gray matter, FWM = frontal white matter, OGM = occipital gray matter, OWM = occipital white matter, cc = corpus callosum.

Determination of Scan Parameters Using Numerical Simulation

To study the signal behavior and determine the appropriate parameters for routine clinical scans at 3.0 T, numerical simulations were performed by solving the Bloch equations and using the relaxation times obtained above. For T1-weighted and T2-weighted sequences, the MR signal for each tissue type is given by:

equation image(1)

with i = gray matter, white matter, or cerebrospinal fluid (CSF).

For the FLAIR sequence, the signal is given by:

equation image(2)

with i = gray matter, white matter, or CSF, where C is the spin density and θ is the flip angle (FA). The factor A denotes the sensitivity of the MR system, which is proportional to the square of the static field strength (12). Thus A3.0T = 4A1.5T.

Image CNR between gray matter and white matter is defined as Scontrast = CNRgw = |SgraySwhite|/Nmath image. Note that the noise level, Nmath image, is proportional to the static field strength (12). Thus N3.0T = 2N1.5T. The noise level was assumed to be independent of the imaging parameters, such as TR, TE, and TI. The optimization criteria for the image contrast are as follows: T1-weighted sequence: Swhite > Sgray > SCSF and to maximize Scontrast; T2-weighted sequence: SCSF > Sgray > Swhite and to maximize Scontrast; and FLAIR sequence: Sgray > SwhiteSCSF (where “≫” indicates more than 10 times greater) and to maximize Scontrast.

The definition of the optimal image contrast is not always trivial. In an MR image, three tissue-specific parameters can influence the image contrast: T1 relaxation time can cause Swhite > Sgray; whereas both T2 relaxation time and spin density effect can cause Sgray > Swhite. Therefore, when choosing imaging parameters for T2-weighted contrast, caution should be taken to ensure that the gray–white matter signal difference is indeed due to T2 relaxation time and not due to spin density. Note that gray matter has a high spin density of 0.88 mL of water/mL of substance, compared to 0.74 mL of water/mL of substance for white matter (13). This generally dominates over the T2 contrast. For instance, when simulations were performed considering both spin density and T2 effects, it was found that the peak Scontrast appeared at a TE of 0 msec. This is because, at longer TE, even though the T2 weighting is gained, which will increase Scontrast, the spin density effect is diminished (CNR down), which causes a more significant decrease in Scontrast. Apparently, this finding of peak TE contradicts the well-established 1.5-T protocol for T2-weighted sequences. However, when identical spin density values were set for gray and white matter, the simulation predicted a peak Scontrast at TE = 90 msec, in excellent agreement with the existing 1.5-T protocols. In view of this, we only considered the T2-related contrast in the simulation for T2-weighted sequence, while forcing the spin density and T1 values of these two tissue types to be the same. The same scheme was used for the FLAIR sequence simulation. However, for the T1-weighted sequence simulation, we found that when only T1 relaxation times were considered, the simulation gave a peak Scontrast at TR = 740 msec at 1.5 T, which is clearly higher than the commonly used values (∼500 msec at 1.5 T). Furthermore, the experimental results acquired at such a TR value showed that the image contrast was not satisfactory (data not shown). This can be explained by the opposing effects of spin density and T1 relaxation on the image contrast, therefore resulting in diminished Scontrast. Thus, simulation for the T1-weighted sequence was performed by considering all MR parameters, including T1, T2, and spin density.

Comparison of Routine Pulse Sequences at 1.5 and 3.0 T

Experiments for comparison between 1.5 and 3.0 T were performed on five normal subjects (three women, two men; age range 25–38 years, mean age 33 ± 6 years) and quantitative analyses of SNR and CNR were conducted. To test the image contrast under various pathological conditions, five patients (three males and two females; age range 31–59 years) were also studied at 3.0 T using the proposed parameters. Two patients had brain tumors (one of them with hemorrhage), one patient had arachnoid cyst, and two patients had incidental white-matter hyperintensities.

For 1.5 T, the standard setup of body-coil transmission and head coil reception was used. For 3.0 T, since the body coil may not be standard equipment at many institutions, a transmit/receive head coil was used instead. For all experiments, 22 axial sections were scanned using the following parameters: FOV = 220 mm; 5-mm slice thickness with 1-mm gap; and matrix size = 256 × 256. Other scan parameters were selected based on the numerical simulation (see Results).

Quantification of SNR and CNR were conducted by the same investigator (L.M.N.P.) using ImageJ. Signals of gray and white matters were determined by selecting an ROI in each tissue type (ROI size: 7.5 ± 0.5 pixels for gray matter and 125.0 ± 12.8 pixels for white matter). Noise level was obtained from the standard deviation (SD) of a region outside the brain in the readout direction. The SNR was then defined as S/Nmath image, and CNR was defined as |SgraySwhite|/Nmath image.

RESULTS

Relaxation Times

Figure 1 shows the absolute T1 and T2 maps for a normal subject at 1.5 T and 3.0 T. Table 1 shows the T1 and T2 values of typical brain structures (see Fig. 1 for examples of ROI selections) at these two field strengths. For all ROIs studied, the T1 values were found to be larger at 3.0 T (14–30%), while T2 values were shorter (12–19%). The largest T1 changes were observed in the genu of the corpus callosum, the largest T2 changes in basal ganglia and thalamus. These results are compared with literature values in Table 2.

Table 1. T1 and T2 Relaxation Times (msec) Obtained at 1.5 T and 3.0 T*
RegionT1_1.5TT1_3TChange (%)aT2_1.5TT2_3TChange (%)a
  • *

    Values are in mean ± standard deviation.

  • a

    Percentage change of values at 3.0 T with respect to that at 1.5 T.

  • FWM = frontal white matter; CC = corpus callosum; OWM = occipital white matter; FGM = frontal gray matter; OGM = occipital gray matter.

FWM556 ± 20699 ± 38+25.779 ± 269 ± 2–12.6
GenuCC556 ± 38721 ± 68+29.776 ± 365 ± 2–14.8
SpleniumCC608 ± 55748 ± 64+23.086 ± 575 ± 5–12.1
OWM616 ± 32758 ± 49+22.992 ± 481 ± 3–11.9
FGM1048 ± 611209 ± 109+15.399 ± 488 ± 3–10.7
OGM989 ± 441122 ± 117+13.590 ± 479 ± 5–12.3
Caudate1009 ± 481258 ± 55+24.685 ± 269 ± 2–18.5
Putamen875 ± 431102 ± 40+26.081 ± 366 ± 2–18.5
Thalamus821 ± 34986 ± 33+20.180 ± 267 ± 2–16.8
Table 2. T1 and T2 Relaxation Times at 3.0 T From the Present Study and From the Literature*
 T1T2
Present studyStudyaStudybStudycPresent studyStudyaStudyd
  • *

    Same abbreviations as in Table 1.

  • a

    Wansapura et al (22).

  • b

    Chen et al (21).

  • c

    Gelman et al (17).

  • d

    Gelman et al (18).

FWM699 ± 38838 ± 18791 ± 27847 ± 4369 ± 274 ± 155.8 ± 3.9
OWM758 ± 49832 ± 18  81 ± 384 ± 1 
GenuCC721 ± 68   65 ± 2  
SpleniumCC748 ± 64   75 ± 5  
FGM1209 ± 1091322 ± 341445 ± 1191763 ± 6088 ± 3110 ± 470.7 ± 10.4
OGM1122 ± 1171283 ± 37  79 ± 5132 ± 9 
Caudate1258 ± 55 1271 ± 911483 ± 4269 ± 2 59.3 ± 4.2
Putamen1102 ± 40  1337 ± 4266 ± 2 51.7 ± 2.2
Thalamus986 ± 33  1218 ± 4067 ± 2  

Comparison of Simulations and Experiments at 3.0 T

T1-Weighted Sequence

Figure 2a shows the simulated gray–white matter contrast, Scontrast, as a function of TR, when using fixed TE = 15 msec and FA = 90°. At 1.5 T, the peak of Scontrast appears at TR = 480 msec, consistent with the values commonly used at this field strength (14). For 3.0 T, the peak was found to be at TR = 520 msec. Figure 3a (left column) shows the images acquired using the respective peak TR at each field strength. Increased contrast between gray and white matter can be seen (arrows) and quantitative analysis indicates that the CNR is improved by 21% at 3.0 T (see Table 3 for details).

Figure 2.

Numerical simulation results of Scontrast = CNRgw as a function of scan parameters. For the T1-weighted sequence, contrast is plotted as a function of TR (a); for the T2-weighted (b) and FLAIR sequences (c), contrast is plotted as a function of TE.

Figure 3.

a: Images obtained from a normal subject at 3.0 T (upper row) and 1.5 T (lower row) using adjusted parameters. From left to right: T1-weighted images, T2-weighted images with fat suppression, and FLAIR images. b: T2-weighted (left column) and FLAIR (right column) images comparing 1-NSA acquisition at 3 T to 2-NSA acquisition at 1.5 T.

Table 3. SNR and CNR Comparison Between 1.5 T and 3.0 T Experimental Data (N = 5)*
Pulse sequenceField strength (T)SNR ± standard errorCNR ± standard error% increase in SNR% increase in SNR (simulation)% increase in CNR% increase in CNR (simulation)
  • *

    Experiments at each field strength were performed using optimal parameters (Table 4) calculated from the simulations.

T1-weighted (SE)1.5    20.734.7
 White matter74.2 ± 5.816.9 ± 2.1    
 Gray matter57.4 ± 4.1     
 3.0      
 White matter101.5 ± 6.820.4 ± 2.736.778.1  
 Gray matter81.1 ± 4.9 41.487.6  
T2-weighted (FSE)1.5    32.496.9
 White matter54.1 ± 3.021.4 ± 0.9    
 Gray matter75.5 ± 2.8     
 3.0      
 White matter82.1 ± 2.928.4 ± 3.951.796.9  
 Gray matter110.5 ± 5.6 46.297.1  
FLAIR1.5    39.339.4
 White matter30.6 ± 1.211.8 ± 0.8    
 Gray matter42.4 ± 1.7     
 3.0      
 White matter47.8 ± 1.416.4 ± 1.056.289.8  
 Gray matter64.1 ± 1.6 51.584.0  
Table 4. Optimal Scan Parameters (Top) and the Adjustable Ranges (Bottom)*
Pulse sequenceField strength (T)TR/TE (msec/msec)TI (msec)NSAScan time
T1-weighted (SE)1.5480/15n/a13 minutes 47 seconds
 3.0520/15n/a13 minutes 36 seconds
T2-weighted (FSE)1.54500/90n/a22 minutes 37 seconds
 3.04500/80n/a22 minutes 37 seconds
FLAIR1.59000/90250024 minutes 39 seconds
 3.09000/80250024 minutes 39 seconds
Pulse sequenceField strength(T)Lower limit (msec)Upper limit (msec)
  1. aDefinition of optimal scan parameters was based on maximum CNR.

  2. bThe adjustable ranges were calculated based on the restriction that 95% of the peak contrast is maintained.

  3. NSA = number of signals averaged.

  4. n/a = not applicable.

T1-weighted (SE) (TR)1.5360620
3.0400660
T2-weighted (FSE) (TE)1.566120
3.058106
FLAIR (TE)1.566120
3.058106

T2-Weighted Sequence

Figure 2b plots Scontrast as a function of TE. The maximum is achieved at TE values of 90 and 80 msec, for 1.5 and 3 T, respectively. Experimental results using such parameters (Fig. 3a; middle column) shows that 3.0 T provides better CNR compared to 1.5 T, yielding an averaged increase of 32% (Table 3).

FLAIR Sequence

The TE dependence of the contrast for a FLAIR sequence is illustrated in Fig. 2c, showing peak Scontrast at 90 and 80 msec, for 1.5 and 3.0T, respectively. Figure 3a (right column) shows the experimental results using the optimal TEs. The CNR at 3.0 T is found to be 39% higher than that at 1.5 T (Table 3).

The scan parameters are summarized in Table 4 (top). To facilitate direct comparison of image contrasts, the scan durations of these experiments were kept constant for both field strengths. However, increased SNR and CNR can also be utilized to shorten scan duration. To demonstrate this possibility, a separate experiment was performed on one normal subject, where the results of 1 NSA at 3.0 T were compared with that of 2 NSA at 1.5 T (Fig. 3b). By visual inspection, it can be seen that the image contrast at 3.0 T is comparable to that at 1.5 T even with 50% time reduction.

Numerical simulations were also performed to determine the upper and lower limits for the imaging parameters, within which 95% of the peak contrast can be achieved (Table 4; bottom). This is useful in providing flexibility for the MR technologists when trying to slightly adjust the brain coverage during the scan. The range of imaging parameters (Table 4; bottom) is quite large for only a 5% CNR variation, due to the fact that the CNR curves (Fig. 2) are flat around the peaks. To verify these recommendations, experiments were performed on a normal subject with a range of scan parameters. The results shown in Fig. 4 indicate that satisfactory contrasts were achieved in all images.

Figure 4.

T1, T2-weighted, and FLAIR images using a range of scan parameters. For each pulse sequence, three different values were used: lower limit for 95% of peak contrast (left), parameter value for peak contrast (middle), and upper limit for 95% of peak contrast (right). Consistent with the simulation results, similar image contrasts were obtained using these parameters. Other scan parameters are the same as the ones used in Fig. 3a.

For FLAIR sequences, the choice of the inversion time is crucial for the CSF signal attenuation. We studied this issue by using three different TIs (2200, 2500, and 2800 msec) and examining the image quality (Fig. 5). It was seen that TI = 2500 msec gave the best CSF nulling, followed by TI = 2200 msec, and the results using TI = 2800 msec were not acceptable. Note that these inversion times were based on a total TR of 9000 msec. Experiments using other TR values should require adjustments for the TI accordingly.

Figure 5.

FLAIR images acquired from a normal subject using different TIs. Note different degrees of CSF nulling (white arrows). The gray–white matter contrast is also slightly different, because the inversion time affects the T1-weighting of the images.

Pathological cases using optimal parameters at 3.0 T showed excellent image contrast and clear lesion delineation. Image qualities at 1.5 and 3.0 T were compared for a patient with incidental white matter hyperintensities (Fig. 6a). Two board-certified neuroradiologists (D.L., M.G.P.) were shown comparable slices of the series without knowledge of field strength or acquisition parameters. Images obtained at 3.0 T were judged to have higher signal and contrast by both readers independently. Despite subtle differences between slice levels for the different systems, the white matter hyperintensities (arrows) were also judged more conspicuous at 3.0 T by both readers. Figure 6b shows 3.0-T images from four other pathological cases (see figure legend for details). The images are all satisfactory as judged by two neuroradiologists (L.M.N.P., D.L.).

Figure 6.

a: 3.0-T and 1.5-T FLAIR images acquired in a patient (42-year-old female) with incidental white matter hyperintensities. 3.0-T and 1.5-T data show equivalent or superior conspicuity of lesions on 3.0 T scans. b: 3.0-T images from four other pathological cases. Case 1: T2W (top), FLAIR (middle), and T1W (bottom) images from a patient (48-year-old male) with an incidental white matter lesion. Case 2: T2W (top), precontrast T1W (middle), and postcontrast T1W (bottom) images from a patient (31-year-old male) with pontine hemangioblastoma. Case 3: FLAIR image from a patient (59-year-old female) with CNS lymphoma and lupus, with left basal ganglia hemorrhage and some periventricular white matter abnormalities. Case 4: Precontrast T1W (top) and postcontrast T1W (middle) images from a patient (53-year-old female) with right cerebellopontine angle arachnoid cyst.

DISCUSSION

With the increasing availability of 3.0-T scanners in research as well as in clinical setups, it is essential to evaluate the impact of this new field strength on the image contrasts of routine clinical sequences. Herein, using numerical simulations and experiments we show that, with appropriate parameter adjustments, clinical images can be obtained in a reasonable scan time with contrast characteristics with which radiologists are acquainted at 1.5 T, yet with higher CNR (21–39%). Furthermore, we show that the image contrast changed minimally (<5%) within a certain range of parameter values and the recommended limits were provided (Table 4).

Behavior of longitudinal relaxation times for different field strengths and different biological tissues can be predicted using relaxation theory (15, 16). However, quantitative measurements using experiments are useful to provide a reference for pulse sequence optimizations. Our results showed that T1 relaxation times were 14% to 30% longer and T2 relaxation times were 12% to 19% shorter at 3.0 T when compared to 1.5 T values, with some regional variations. Interregional brain variation of T1 relaxation times at 3.0 T is often correlated to nonheme iron levels and water content; iron levels have greater influence on gray matter relaxation times, whereas water content has greater influence on white matter (17). T2 relaxation times are also highly correlated to iron and water content (18) as well as to cytoarchitecture (19). Our results showed the largest variation in T2 relaxation times for the caudate nucleus and putamen, regions rich in iron content, exhibiting the lowest values at high magnetic field. For CSF, we found no differences in T1 values between 1.5 T (3836 ± 470 msec) and 3.0 T (3817 ± 424 msec; N = 10; P > 0.05). However, the range of the values is lower than the values commonly used for FLAIR at 1.5 T, which can be calculated to be 4357 msec using a TR of 9000 msec and a CSF nulling TI of 2500 msec. This is probably because our T1 measurement sequence used a recovery time of 8000 msec, which is not long enough for accurate determination of CSF T1. Therefore, we used CSF literature values (4300 msec for both field strengths) (20, 21) for FLAIR sequence simulation.

Compared to previously reported values at 3.0 T (Table 2) (17, 18, 21, 22), our relaxation times lie in the low range for T1 and in the middle range for T2 relaxation times. A small variation is expected, as measurements differ depending on the region chosen for analysis, spatial resolution (e.g., due to CSF partial volume effects in gray matter). However, some differences are systematic and are related to the pulse sequences used for measurement. For instance, Wansapura et al (22) reported gray matter T2 relaxation time of 110—132 msec at 3.0 T, which is about 20% to 40% higher than our results. This can be explained by the shorter interecho spacing (τCPMG = 10 msec) used in their study. T2 relaxation times are known to be dependent on the interecho spacing due to the water diffusion in an inhomogeneous magnetic field and due to exchange contributions (23). Shorter τCPMG usually results in longer T2 values, and such effect is more prominent in structures containing significant amounts of blood (24) or regions with large magnetic susceptibility gradients, e.g., at tissue surfaces. However, it should be noted that when considering signals in conventional SE T2-weighted images, the signal behavior should be based on the T2 at a τCPMG of TE (80–90 msec). Therefore, we believe that it is more appropriate to use T2 measured at long τCPMG in the simulations.

In our T2 measurement, a segmented EPI acquisition scheme was used to reduce the scan duration. As a result, the T2 weighting in the MR signal is expected to be an average of a range of TE values and have slight Tmath image character. However, because the echo-train is symmetric around the nominal TE, this is expected to have a minimal effect on the calculated T2 maps.

Imaging parameters at 1.5 T vary slightly for different institutions, but are within similar ranges. The recommended 1.5-T parameters from this study (Table 4) are in general agreement with recommended values from major MR vendors (GE Healthcare, Philips Medical Systems, Siemens Medical Solutions), which are as follows: T1-weighted: TR = 400–600 msec, FA = 90°, TE = minimum (∼15 msec); T2-weighted: TR = 3000–6000 msec, FA = 90°, TE = 90–120 msec; FLAIR: TR = 6000–11,000 msec, TI = 2000–2800 msec, FA = 90°, TE = 90–140 msec (values obtained from three different institutions with different MR vendors). Our 1.5-T parameters are also consistent with values in the literature (14).

In our quantitative assessment of the image qualities, each sequence was optimized by focusing on its most important scan parameters, namely, TR for T1-weighted, TE for T2-weighted, and TI and TE for FLAIR sequences. The effects of other scan parameters were also studied qualitatively (N = 1, 3.0 T). For the T1-weighted sequence, experiments were performed at different TEs, 11 msec (shortest TE achievable without changing other parameters), 15, and 20 msec, and the resulting images showed very similar appearances (data not shown). Therefore, we believe that, in a T1-weighted sequence, TE has a minor effect on the image contrast when it is kept to less than 20 msec. For T2-weighted sequence, more than 98% of the equilibrium longitudinal magnetizations have recovered at the TR used in this study (4500 msec), and the gray–white matter difference was less than 1.2%. Thus, the remaining T1 weighting is negligible when using a TR of 4500 msec.

The 3.0-T results showed higher SNR and CNR for all sequences tested. However, it can be seen (Table 3; last four columns) that the SNR and CNR increases in the experimental data are smaller than the values expected from the simulations. While the situation for CNR varies considerably depending on the individual sequence, the SNR mismatch between experiments and simulations appears to be consistent among the three sequences tested. The ratio between simulation and experimental results are 1.3, 1.3, and 1.2 for T1-weighted, T2-weighted, and FLAIR sequences, respectively. It can also be calculated that the ratio of overall sensitivity between the 3.0-T and the 1.5-T system was 1.56, which is lower than the theoretical value of 2.0 (12). One possible reason for such discrepancy is that a body coil was used for transmitting at 1.5 T, whereas at 3.0 T a head coil was used, which is known to have poorer transmission efficiency and B1 homogeneity.

Increases in static field strength will also increase the energy deposition of the RF pulses. When using the same pulse power and length, the specific absorption rate (SAR) will become four times higher for 3.0 T and requirements to minimize this increase may affect the scan duration and image contrast. For our 1.5-T experiments, in which a body coil was used and the peak B1 field was 27.1 μT, the SAR-values were 1.4, 2.9, and 1.9 W/kg for T1-weighted, T2-weighted, and FLAIR sequences, respectively. For the 3.0-T sequences, if a body coil were used with an identical peak B1 field, the SAR for the T2-weighted sequence would have been 7.2 W/kg, which is considerably higher than the SAR limit (3.0 W/kg). One can, of course, alleviate the SAR problem by increasing TR, decreasing the number of RF pulses, decreasing number of slices, etc. An alternative approach is to use a transmit/receive head coil and/or to decrease the peak B1 field. The transmit/receive head coil delivers significantly less energy due to decreased body coverage. The B1 field can be adjusted because the flip angle is linearly proportional to the B1 field, whereas the energy deposition is quadratically proportional to B1. Therefore, by using a smaller B1 with longer pulse duration, the same flip angle is achieved but the energy is decreased considerably. It was found that for a T2-weighted sequence at 3.0 T, a SAR comparable to the one at 1.5 T (2.9 W/kg) can be achieved by using a body coil with a peak B1 field of 11.1 μT or by using a transmit/receive head coil with a peak B1 of 14.3 μT. Similar calculations can be carried out for T1-weighted and FLAIR sequences. In this study, the protocols used at 3.0 T were based on a transmit/receive head coil with a peak B1 of 10.3 μT, resulting in SAR values of 1.0, 2.1, and 1.4 W/kg for T1-weighted, T2-weighted, and FLAIR sequences, respectively. Note that with the availability of parallel imaging techniques, the body coil can now be used for routine clinical scans without increasing scan duration, since the parallel imaging allows one to reduce the number of k-lines acquired (25).

In some institutions, the diffusion weighted imaging (DWI) sequence is also routinely performed for brain MRI scans. Since the diffusion coefficient is a physical/physiological parameter of the tissue and is not field-dependent, we expect a minimal change in scan parameters when comparing the 3.0-T to the 1.5-T protocol. However, since DWI is also T2-weighted, the shortening of T2 at 3.0 T may cause a slight change in image contrast. For systems in which higher gradient strengths are available, this can be corrected by reducing TE while keeping the b-value the same. The SNR of the DWI images at 3.0 T will be higher than that at 1.5 T, although the improvement will be less than the theoretical value of 100% due to faster T2 relaxation at the higher field.

Although qualitative evaluation of images carries a high degree of subjectivity, it reflects the daily activity in radiology practice. In addition to the expected greater signal and contrast for images acquired at the higher field, some characteristics specific to the images obtained at 3.0 T were observed. It is important for radiologists to recognize these findings that are intrinsic to the higher field and not to overcall them as abnormalities. For example, T2-weighted images acquired at 3.0 T accentuate the high iron content structures, including the globus pallidus, red nuclei, substantia nigra, and dentate nuclei. Susceptibility artifacts are known to be more accentuated at 3.0 T. In addition, previous reports have shown better lesion detection on T1-weighted images after gadolinium injection at the higher field (5, 6, 8).

In conclusion, measurements of relaxation times of major brain structures showed increased T1 values and decreased T2 values at 3.0 T, which can significantly affect the MR image contrast. However, with recommended adjustment of the imaging parameters, the routine brain sequences (T1-weighted, T2-weighted, and FLAIR) can show comparable contrast to that at 1.5 T but with higher CNR. High field strength provides the potential to achieve higher spatial resolution and/or to shorten the scan duration. We believe that imaging at 3.0 T is advantageous over that at 1.5 T for routine clinical brain scans.

Acknowledgements

We thank Kathie Kahl, Terri Brawner, Joe Gillen, Steve Ringold, and Scott Pryde for experimental assistance. Dr. van Zijl is a paid lecturer for Philips Medical Systems. This arrangement has been approved by Johns Hopkins University in accordance with its conflict of interest policies. Supported by National Institutes of Health (NIH)/NIBIB grant R21 EB00991-01 and NIH/NCRR grant P41-RR15241 (both to P.vZ.).

Ancillary