Relaxation Times of Breast Tissue at 1.5T and 3T
Measured Using IDEAL
Rebecca Rakow-Penner, MS, Bruce Daniel, MD, Huanzhou Yu, MS,
Anne Sawyer-Glover, BS, RT, and Gary H. Glover, PhD*
Purpose: To accurately measure T1 and T2 of breast fi-
broglandular tissue and fat at 1.5T and 3T, and note the
partial volume effects of the admixture of fibroglandular
tissue and fat on the relaxation rates using an approach
termed iterative decomposition of water and fat with echo
asymmetry and least squares estimation (IDEAL) imaging.
Materials and Methods: T1 and T2 values were measured
on the right breasts of five healthy women at 1.5T and 3T.
T1 data were collected using two sequences: inversion re-
covery without IDEAL, and inversion recovery with IDEAL.
T2 data were collected using Hahn Echo scans. SNR and
CNR analyses were conducted on collected data.
Results: T1 increased for both fat (21%) and glandular
tissue (17%) from 1.5T to 3T. Thus, the TR and TI of breast
protocols at 3T should be lengthened accordingly. SNR
more than doubled for both tissue types from 1.5T to 3T.
IDEAL imaging demonstrated the partial volume effects of
fat and glandular tissue on measuring relaxation rates of
independent tissue types.
Conclusion: With separated fat and water images, more
precise measurements can be made for the lipid component
in fat, and the water component in fibroglandular tissue.
Key Words: relaxation times; breast; MRI; IDEAL; fat/water
J. Magn. Reson. Imaging 2006;23:87–91.
© 2005 Wiley-Liss, Inc.
THE HIGHER SENSITIVITY of magnetic resonance im-
aging (MRI) compared to x-ray mammography has led
to growing clinical interest in the use of contrast-en-
hanced MRI of breast tissue, especially for women at
high risk for breast cancer (1). Dynamic breast MRI
depends on the sensitivity and specificity of the imaging
protocol used, specifically the temporal and spatial res-
olution (2–4). With an increased signal-to-noise ratio
(SNR), the resultant higher spatial resolution and faster
acquisition times may improve sensitivity and specific-
ity for morphologic and dynamic breast MRI tech-
niques. Doubling the field strength, such as imaging at
3T vs. 1.5T, approximately doubles the SNR, albeit at
an increase in T1 (5). Imaging at 3T thus holds valuable
potential for refining diagnostic breast MRI.
In transitioning breast imaging protocols from 1.5T to
3T, the scan parameters should be adapted to optimize
contrast and SNR at the increased field strength. Gen-
erally, higher field strength increases the longitudinal
relaxation time (T1) and slightly decreases the trans-
verse relaxation time (T2) of breast tissues (6). In defin-
ing a scan protocol for a tissue of interest, the relaxation
time (TR) should increase as T1 increases; otherwise,
magnetization saturation effects will diminish the SNR
gain (7,8). Thus the TR should be adjusted according to
the T1 for the tissue of interest. Breast tissue is primar-
ily composed of glandular tissue and lipids. To the best
of our knowledge, there are currently no published val-
ues for T1 and T2 of breast glandular tissue and fat at
3T. However, subcutaneous fat has been measured (9)
and provides a guiding value for breast fat T1 and T2
Relaxation times for breast tissue have been mea-
sured at field strengths of ?1.5T (10–14). The pub-
lished values often contradict one another, possibly be-
equipment, and varying admixtures of fat and water in
breast tissue. The relaxation rates of other human tis-
sues, such as musculoskeletal tissue (15), abdominal
and pelvic tissue (9), and brain (16), have been mea-
sured at 3T. These studies indicated that the SNR and
T1 of the respective measured tissues increased from
1.5T to 3T. They also noted a minimal change of T2 with
increasing field strength.
In the present study we measured the T1 and T2
values of breast fibroglandular tissue and fat at 1.5T
and 3T, and performed a contrast-to-noise ratio (CNR)
analysis of breast imaging. Importantly, this study
notes the partial volume effects of the admixture of fat
and glandular tissue in breasts when the independent
relaxation rates of the respective tissues are measured.
These partial volume effects result from the biological
water component in fat, and the lipid component in
glandular tissue, as well as breast architecture. To
Department of Radiology, School of Medicine, Stanford University,
Stanford, California, USA.
Contract grant sponsor: National Institutes of Health (NIH); Contract
grant number: P41-RR09784.
*Address reprint requests to: G.H.G., Stanford University School of
Medicine, Lucas MR Center, 1201 Welch Road, Stanford, CA 94305-
5488. E-mail: email@example.com.
Received January 3, 2005; Accepted September 27, 2005.
Published online 28 November 2005 in Wiley InterScience (www.
JOURNAL OF MAGNETIC RESONANCE IMAGING 23:87–91 (2006)
© 2005 Wiley-Liss, Inc.
eliminate these effects we used an approach termed
iterative decomposition of water and fat with echo
asymmetry and least squares estimation (IDEAL)
(17,18), which provides separated fat and water images
on which to base the T1 and T2 measurements. IDEAL
uses three measurements with different TEs to sepa-
rate fat and water. It is robust to B0 inhomogeneities,
since a map of the field strength is calculated as part of
the reconstruction (18).
MATERIALS AND METHODS
Five healthy women, ranging in age between 24 and 51
years (mean age ? 36.4 years, standard deviation
(SD) ? 12.6), participated in this study after they gave
informed consent according to a protocol approved by
the local IRB. The study consisted of three scan se-
quences performed on the right breast of each volunteer
at 1.5T and 3T (Echospeed whole-body magnet; GE
Healthcare, Waukesha, WI, USA): 1) T1 data collection
using fast spin-echo inversion recovery (FSE-IR) for
combined fat and water images, 2) T2 data collection
using Hahn echo acquisition, and 3) T1 data collection
using FSE-IR IDEAL for separated fat and water im-
ages. The FSE-IR IDEAL technique provides a com-
bined image as well. However, the IDEAL combined
image results in a lower SNR than its separated fat and
water images, and thus was not used for any calcula-
tions. The SNR of the combined image is diminished
because of correlation between the fat and water images
(20). A non-IDEAL FSE-IR was used to measure a com-
bined fat and water image. Note that there is some bias
toward the IDEAL separated images with regard to SNR
because the reconstructions are an outcome of three
excitations, whereas the non-IDEAL image is an out-
come of one excitation. The FSE-IR and Hahn se-
quences are commercially available. The FSE-IR IDEAL
sequence is an in-house-developed sequence. This
technique was preferred for T1 measurements over
other, faster techniques because of its robustness and
dependability. Since the volunteers were all healthy
and no contrast was injected, time was not a limiting
factor. If one wanted to perform T1 measurements on
patients or a large population, a faster but potentially
less reliable technique might be preferable.
Four-channel receive-only breast coils (Invivo/MRI
Devices Corp., Waukesha, WI, USA) were used to ac-
quire the data. Each volunteer was scanned at both
field strengths no more than 2 days apart to control for
menstrual cycle variations in the breast tissue within
(but not among) individuals.
The FSE-IR protocol for T1 measurements consisted
of four scans with variable TIs of 3000, 800, 300, and
150 msec, and the following parameters: TE ? 18.5
msec, TR ? 4000 msec, echo train length ? 8, band-
width ? 15.63 kHz, FOV ? 20 mm ? 20 mm, slice
thickness ? 5 mm, slice direction ? sagittal, NEX ?
1, resolution ? 256 ? 128. An adiabatic inversion
pulse was used in the vendor-supplied pulse se-
quence. Only four TIs were collected, due to time
constraints. To calculate T1, we assumed monoexpo-
nential decay. The following equation was solved for
T1, given the measurements Micorresponding to the
Mi? M0?1 ? 2?1 ? ε?exp??TIi/T1??,(1)
where M0is the equilibrium magnetization level, and ε
is an error term that accounts for imperfect magnetiza-
tion inversion by the 180° pulse (ε was nominally close
to zero, but no matter how small it was it remained in
the calculation). The equation was solved using the
Nelder-Mead simplex method as implemented by MAT-
LAB (Natick, MA, USA). When the T1 for fat was calcu-
lated, the received signal, M, resulted from an average
of three regions of interest (ROIs) drawn in the fat region
for each TI value per volunteer (see Fig. 1). The size of
the ROIs varied because they were drawn to maximize a
region of homogeneous tissue type. After each volun-
teer’s T1 value in fat was solved for, the final value
resulted from averaging T1 values across all volunteers
for a given field strength. The same method was used to
calculate the T1 of glandular tissue. The order of aver-
aging the signals and then estimating the parameters
per volunteer vs. estimating the parameters per volun-
teer and then averaging the parameters did not produce
significantly different results.
The protocol for the T2 scans consisted of two Hahn
echo scans with two TEs (TE1 and TE2) of 20 and 100
msec, respectively, and the following scan parameters:
TR ? 2000 msec, bandwidth of 15.63, FOV ? 20 mm,
slice thickness ? 5 mm, slice direction ? sagittal,
NEX ? 1, resolution ? 256 ? 128. Assuming a mono-
exponential T2 decay, the transverse magnetization Mi
for TEiis given by:
Figure 1. The white dots represent sample fat ROIs, and the
black dots represent sample glandular-tissue ROIs on an
IDEAL combined image.
88Rakow-Penner et al.
where M0is a constant. T2 and M0can be solved with
the two known measurements Mi(TEi). We measured M
for fat and glandular tissue by placing three ROIs in the
fat and glandular-tissue regions, respectively, of the
reconstructed image. The fat and water ROIs were av-
eraged independently per volunteer. As in the T1 calcu-
lation, the overall T2 values for each fat and water ROI
resulted from an average across volunteers per field
The parameters of the IDEAL sequence remained the
same as for the non-IDEAL FSE-IR protocol. The IDEAL
technique produces three sets of images: combined fat
and water, separated water, and separated fat (18) (see
Fig. 2). Because of the above-mentioned SNR disadvan-
tage, we did not use the IDEAL combined image. For the
separated images, ROIs outside of the fat regions in the
fat images, and outside glandular-tissue regions in the
water images were not measured because it proved dif-
ficult to draw an ROI of fat in water and vice versa
consistently across volunteers. Thus, only three ROIs
were drawn in each separated image in the specified
tissue type. As in the non-IDEAL analysis, the size of
the ROIs depended on a maximized area of the homo-
geneous tissue of interest. The ROIs drawn in the
IDEAL and non-IDEAL analyses differed slightly be-
cause the acquisitions differed, although the ROIs in
both acquisitions were approximately in the same po-
sition and of the same size. The calculation process
followed the previous approach of solving T1 for each
volunteer in the separated fat and water images, and
then averaging over all the volunteers.
SNR and CNR calculations were performed on the
FSE-IR T1 data. The signal was taken to be the average
M0values (fully relaxed magnetization) in the two tissue
types for each subject at each field. The noise was mea-
sured with an ROI placed in the M0image corner, with
care taken to avoid ghosting artifacts. No correction
was made for the Rician distribution of the magnitude
images. CNR measurements were obtained between fat
and water using three specified ROIs (fat, glandular
tissue, and noise) placed in FSE-IR images at each of
the four TIs.
Increasing field strength led to an increase in T1 values
for both fat and water (refer to Table 1). The SD analysis
was performed between subjects. IDEAL imaging pro-
vided lower values for T1 in fat, and higher values for T1
in water at both 1.5T and 3T. Significant differences
between T1 values were examined with t-tests for fat
non-IDEAL and IDEAL at 1.5T, water non-IDEAL and
IDEAL at 1.5T, fat non-IDEAL and IDEAL at 3T, and
water non-IDEAL and IDEAL at 3T (Table 1). These
significant P-values demonstrate a trend in the impact
of the admixture of fat and water in both fatty and
fibroglandular tissue. Note that the SDs for the IDEAL
technique are smaller than those for the non-IDEAL
approach. This may partially result from the bias of the
3 NEX IDEAL vs. the 1 NEX non-IDEAL.
T2 values (refer to Table 2) did not demonstrate a
significant difference between field strengths of either
tissue. Also, T2 values did not significantly differ be-
tween tissue types at either field strength. Since these
Figure 2. IDEAL images at 1.5T with a
TI of 3000 msec: (a) combined fat and
water, (b) fat, and (c) water. Note that the
water component of the fatty tissue is
visible in slice c.
T1 Relaxation Times at 1.5T and 3T†
T1 (average ? SD,
T1 (average ? SD,
372.04 ? 8.6
1135.98 ? 151.37
296.01 ? 12.94*
1266.18 ? 81.8*
449.27 ? 26.09**
1324.42 ? 167.63**
366.78 ? 7.75**
1444.83 ? 92.7**
FSE-IR with IDEAL
†The standard deviation listed is for the deviation between T1 values.
*P ? 0.05 between non-IDEAL vs. IDEAL with the same field strength.
**P ? 0.05 between 1.5T and 3T of the same tissue type and scan parameters.
Relaxation Times of Breast Tissue 89
results showed little difference, IDEAL results were not
The ratio of the fat/water CNR values at 3T and 1.5T
without IDEAL was 2.18. The CNR ratio between 3T and
1.5T with IDEAL was 3.15.
The results indicate that with increasing field strength,
from 1.5T to 3T the T1s of fat and fibroglandular tissue
increased by 21% and 17%, respectively, using the non-
IDEAL technique. With IDEAL imaging the T1s of fat
and fibroglandular tissue increased by 23% and 14%,
respectively. The parameters used for scanning at 3T
must be optimized accordingly. In particular, TR and TI
values should be lengthened at 3T. According to Edel-
stein et al (19), the TR for saturation recovery imaging,
and the TI for inversion recovery imaging should be set
approximately equal to the T1 of the tissue of interest at
a given field strength.
Our novel approach for measuring T1 of fat and
glandular tissues from IDEAL separated images pro-
vides greater precision in T1 measurements by reduc-
ing partial volume effects. For measurements of the
lipid component of fat or the water component of
glandular tissue, optimizing scan protocols based on
values obtained from IDEAL T1 images will provide
more exact results. When the combined IDEAL image
results (Fig. 2a) are compared with the IDEAL image
results in which fat and water are separated (Fig. 2b
and c), the fat T1 values across field strengths drop
notably. This is because breast fat has both water
and lipid components. The values measured in the
IDEAL separated fat image provide only the T1 values
of the lipid component. Thus, to accurately measure
the T1 of the lipid component in breast tissue, IDEAL
separation is recommended. On the other hand, T1s
of fibroglandular tissue in the IDEAL water separated
images are significantly higher than in the combined
measurements, indicating the value of separating the
water and lipid components when examining the wa-
ter content in breast tissue.
T2, as measured with a spin-echo sequence, does not
significantly vary between 1.5T and 3T (it varies less
than 5 msec), nor does it vary significantly between fat
and glandular tissue. Thus the TE does not require
A SNR analysis demonstrated the predictable linear
upward relationship between SNR and field strength.
CNR calculations confirmed that T1 tissue contrast
improves with increasing field strength. This en-
hancement more than compensates for any loss in
saturation recovery due to higher magnetic fields.
This paper focused on the SNR and CNR of fat and
glandular tissue without contrast agents. For diag-
nostic purposes, measuring these ratios with con-
trast agents would be useful and may be done in
The results of this study can serve as a guideline for
adjusting breast protocols for 3T imaging, and spe-
cifically indicate that the TR must be lengthened pro-
portionally to the T1 of the breast tissue of interest.
Our study had the following limitations: 1) a small
sample size was used, 2) data were collected from
healthy volunteers vs. cancer patients, and 3) men-
strual variations were controlled for within but not
among the volunteers. The results from our limited
sample group provide general protocol guidelines for
glandular and fat tissue. A larger sample group may
provide more statistically accurate results. Because
of the length of the scans and the repetition of acqui-
sitions at two field strengths, data were collected on
healthy volunteers rather than cancer patients. T1
data on lesions are useful and may be included in
future studies. Our study controlled for menstrual
variations within volunteers by scanning them at
both field strengths within a short span of time. Con-
trolling for menstrual variations among volunteers
would have required greater flexibility of magnet
time, and inconvenienced the volunteers.
In conclusion, T1 measurements in the breast with
IDEAL imaging provide more precise values for fat and
glandular tissue, and reduce patient-to-patient vari-
ability in the values measured. The TRs and TIs of
breast protocols at 3T must be lengthened from the
values used at 1.5T to account for the increase in T1 of
We thank Drs. Kim Butts and Karl Vigen for their input
on this project, and Ann Shimakawa and Dr. Scott
Reeder for providing the FSE-IR IDEAL sequence.
1. Kriege M, Brekelmans C, Boetes C, et al. Efficacy of MRI and mam-
mography for breast-cancer screening in women with a familial or
genetic predisposition. N Engl J Med 2004;351:427–437.
2. Agoston AT, Daniel BL, Herfkens RJ, et al. Intensity-modulated
parametric mapping for simultaneous display of rapid dynamic and
high-spatial-resolution breast MR imaging data. Radiographics
3. Kuhl CK, Bieling HB, Gieseke J, et al. Healthy premenopausal
breast parenchyma in dynamic contrast-enhanced MR imaging of
the breast: normal contrast medium enhancement and cyclical-
phase dependency. Radiology 1997;203:137–144.
4. Song HK, Dougherty L, Schnall MD. Simultaneous acquisition of
multiple resolution images for dynamic contrast enhanced imaging
of the breast. Magn Reson Med 2001;46:503–509.
5. Haacke EM, Brown RW, Thompson MR, et al. Magnetic resonance
imaging: physical principles and sequence design. New York: John
Wiley and Sons; 1999. p 1–914.
T2 Relaxation Times at 1.5T and 3T*
T2 (msec) SDT2 (msec) SD
*There was no significant difference between T2 values due to tissue
types or field strength.
90 Rakow-Penner et al.
6. Bottomley PA, Foster TH, Argersinger RE, et al. A review of normal
tissue hydrogen NMR relaxation times and relaxation mechanisms
from 1–100 MHz: dependence on tissue type, NMR frequency, tem-
peratures, species, excision and age. Med Phys 1984;11:425–448.
7. Jezzard P, Duewell S, Balaband RS. MR relaxation times in human
brain: measurement at 4T. Radiology 1996;199;773–779.
8. Crooks LE, Arakawa M, Hoenninger J, et al. Magnetic resonance
imaging: effects of magnetic field strength. Radiology 1984;151:
9. de Bazelaire CMJ, Duhamel GD, Rofsky NM, et al. MR imaging
relaxation times of abdominal and pelvic tissues measured in vivo
at 3.0T: preliminary results. Radiology 2004;230:652–659.
10. Bottomley PA, Hardy CJ, Argersinger RE, et al. A review of H
nuclear magnetic resonance relaxation in pathology: are T1 and T2
diagnostic? Med Phys 1987;14:1–37.
11. Heywang SH, Fenzl G, Hah Dietbert, et al. MR imaging of the
breast: comparison with mammography and ultrasound. J Comput
Assist Tomogr 1986;10:615–620.
12. Partain CL, Kulkarni MV, Price RR, et al. Magnetic resonance im-
aging of the breast: functional T1 and three-dimensional imaging.
Cardiovasc Intervent Radiol 1986;8:292–299.
13. El Yousef SJ, O’Connell DM, Duchesneau RH, et al. Benign and
malignant breast disease: magnetic resonance and radiofrequency
pulse sequences. AJR Am J Roentgenol 1985;145:1–8.
14. Merchant TE, Thelissen GRP, de Graaf PW, et al. Application of a
mixed imaging sequence for MR imaging characterization of human
breast disease. Acta Radiol 1993;34:356–361.
15. Gold GE, Han E, Stainsby J, et al. Musculoskeletal MRI at 3.0T:
relaxation times and image contrast. AJR Am J Roentgenol 2004;
16. Wansapura JP, Holland SK, Dunn RS, et al. NMR relaxation times
in the human brain at 3.0 Tesla. J Magn Reson Imaging 1999;9:
17. Glover GH, Schneider E. Three-point Dixon technique for true
water/fat decomposition of B0 inhomogeneity correction. Magn
Reson Med 1991;18:371–383.
18. Reeder SB, Pineda AR, Wen Z, et al. Iterative decomposition of
water and fat with echo asymmetry and least squares estimation
(IDEAL): application with fast spin-echo imaging. Magn Reson Med
19. Edelstein WA, Bottomley PA, Hart HR, et al. Signal, noise, and
contrast in nuclear magnetic resonance (NMR) imaging. J Comput
Assist Tomogr 1983;7:391–401.
20. Yu H, Reeder SB, Shimakawa A, Gold GE, Pelc NJ, Brittain JH.
Implementation and noise analysis of chemical shift correction for
fast spin echo Dixon imaging. In: Proceedings of the 12th Annual
Meeting of ISMRM, Kyoto, Japan, 2004. p 2686.
Relaxation Times of Breast Tissue91