Note: This copy is for your personal non-commercial use only.

To order presentation-ready
copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

FUTURE DIRECTIONS S241

Proton (1H) MR Spec-

troscopy of the Breast1
Lia Bartella, MD ● Wei Huang, PhD
TEACHING
POINTS
See last page Proton (hydrogen 1) [1H]) magnetic resonance (MR) spectroscopy
provides biochemical information about the tissue under investigation.
Its diagnostic value in cancer is typically based on the detection of el-
evated levels of choline compounds, choline being a marker of active
tumor. The two main potential clinical applications of 1H MR spec-
troscopy are (a) as an adjunct to breast MR imaging to improve speci-
ficity in differentiating benign from malignant lesions, and (b) for
monitoring or even predicting response to treatment in patients under-
going neoadjuvant chemotherapy. Preliminary data are promising, with
study results suggesting that 1H MR spectroscopy may decrease the
number of benign biopsies recommended on the basis of MR imaging
findings and may help predict response as early as 24 hours after the
first dose of neoadjuvant chemotherapy. Although several limitations
currently exist that make the technique premature for clinical use, fur-
ther evaluation with larger, preferably multicenter trials is certainly
warranted.
©
RSNA, 2007

Abbreviations: DCIS ⫽ ductal carcinoma in situ, PPV ⫽ positive predictive value, SNR ⫽ signal-to-noise ratio

RadioGraphics 2007; 27:S241–S252 ● Published online 10.1148/rg.27si075504 ● Content Codes:

1From the Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, New York, NY 10021. Received February 5, 2007;
revision requested March 15 and received May 16; accepted June 21. Both authors have no financial relationships to disclose. Address correspon-
dence to L.B. (e-mail: liabartella@hotmail.com).
©
RSNA, 2007
 S242 October 2007 RG f Volume 27 ● Special Issue

Introduction ated by mobile lipids, which sidebands can ob-

Proton (hydrogen 1) [1H]) magnetic resonance scure the detection of the choline peak. This lipid
(MR) spectroscopy is not a new technology, hav- signal problem arises because of the large amount
ing been used to provide biochemical information of fat in the breast. Typical acquisition parame-
about biologic tissues for over 30 years. 1H MR ters include an echo time of 135 msec or longer to
spectroscopy has been approved by the U.S. reduce lipid signal and a repetition time of 1.5–
Food and Drug Administration and is widely 3.0 seconds. The number of signals acquired is
used for evaluation of the brain and prostate usually between 128 and 256, resulting in a net
gland. Overall, however, its evolution has been data acquisition time of 3.2–12.8 minutes. An
slow in the clinical setting and even slower in extra 5–10 minutes is needed for preacquisition
breast studies. Although quantification of me- set-up of 1H MR spectroscopy voxel shimming
tabolite levels is routinely performed in 1H MR and water suppression. In cases of multiple sus-
spectroscopy of the brain, it is more difficult to pect lesions in one breast, multivoxel MR spec-
perform in breast evaluation because of the hetero- troscopy is the preferred technique. This tech-
geneous distribution of the glandular and adipose nique provides information about the spatial dis-
tissues (1). With respect to the breast, 1H MR tribution of metabolites and is useful for studying
spectroscopy is still an investigational technique multiple lesions. It can be used to measure multi-
with a promising future in the clinical setting. regional metabolite levels in a data acquisition
The diagnostic value of 1H MR spectroscopy time comparable to that for a single-voxel study.
Teaching in cancer is typically based on the detection of However, mostly due to difficulty in obtaining
Point elevated levels of choline compounds, choline good shimming in a relatively large breast region
being a marker of active tumor (2). These com- within a reasonable time frame, only a few studies
pounds have methyl protons that resonate at a have yielded data of acceptable quality (5,6). For
chemical shift of 3.2 ppm (3). The composite differentiating benign from malignant breast le-
resonance at 3.2 ppm includes contributions from sions, except for a study in which choline concen-
choline, phosphocholine, glycerophosphocholine, tration was quantified using water signal as the
myoinositol, and taurine (4). internal reference (1), the majority of 1H MR
In this article, we review 1H MR spectroscopic spectroscopic studies are based on detection (or
technique; discuss and illustrate the potential nondetection) of the choline peak or its signal-to-
clinical applications of this modality in the breast noise ratio (SNR) (7).
(differentiating benign from malignant breast le- Breast 1H MR spectroscopy has several draw-
sions, predicting response to neoadjuvant chemo- backs. Prior contrast material– enhanced MR im-
therapy); and describe the current limitations of aging is usually required for lesion localization
1H MR spectroscopy in this setting. and MR spectroscopic voxel placement. The ac-
cumulation of contrast agent in the lesion can
Technical Considerations affect 1H MR spectroscopic quality due to T2*
Breast 1H MR spectroscopy is usually performed broadening effect (8). Also, the time required (in-
on a clinical magnet with a field strength of at cluding preacquisition adjustment) is relatively
least 1.5 T. A breast coil is also needed, just as for long (10 –25 minutes) and the spatial resolution
MR imaging. Spectroscopic sequences are com- poor. Fine tumor heterogeneity cannot be as-
mercially available, but at the present time, a sayed. It is difficult to achieve sufficient simulta-
spectroscopist needs to be involved to perform neous suppression of water and lipid resonances,
off-line data processing. making it difficult to quantify choline concentra-
Breast 1H MR spectroscopy has predominantly tion. Thus, the majority of 1H MR spectroscopic
been performed with a single-voxel technique. studies are non- or semiquantitative. Further-
This technique is limited to evaluating one lesion more, because of the difficulty of detecting weak
at a time. A voxel is placed to encompass the le- choline signal from a small lesion within a reason-
sion or area of interest. In most single-voxel 1H able time frame in a clinical setting at 1.5 T, the
MR spectroscopic studies of the breast, the point- sensitivity of 1H MR spectroscopy in detecting
resolved spectroscopic sequence or a variation breast malignancy drops dramatically when the
thereof is used for data acquisition. One such lesion is less than 2 cm in diameter (9). With
variation is the incorporation of echo time averag- expected improvement in SNR, higher-field-
ing technique into a regular point-resolved spec- strength (eg, 3-T) MR imagers may allow 1H MR
troscopic sequence (1). This method reduces the spectroscopic investigation of smaller lesions with
sidebands that result from spurious echoes gener- high sensitivity within a reasonable time frame.
Higher field strength will also improve spectral
resolution, which means better separation be-
tween water, choline, and fat peaks. This im-
RG f Volume 27 ● Special Issue Bartella and Huang S243

Figure 1. (a) Sagittal non-fat-suppressed T1-weighted MR image (repetition time msec/

echo time msec ⫽ 6.4/3.1) of the right breast, obtained in a 65-year-old woman with a his-
tory of pseudoangiomatous stromal hyperplasia in the left breast, shows normal glandular
parenchyma and fat. (b) Magnified spectrum illustrates a high lipid (Lip) peak, but no cho-
line (Cho) resonance peak is observed at a frequency of 3.2 ppm. Lac ⫽ lactate.

Figure 2. (a) Sagittal non-fat-suppressed T1-weighted MR image (6.4/3.1) of the left

breast, obtained in a 47-year-old woman during week 2 of the menstrual cycle, shows nor-
mal glandular parenchyma and little fat. The patient had undergone lumpectomy in the con-
tralateral breast for cancer discovered at screening breast MR imaging performed owing to
the patient’s high risk. (b) Magnified spectrum illustrates a high lipid (Lip) peak, but no
choline (Cho) resonance peak is observed at a frequency of 3.2 ppm. Lac ⫽ lactate.

proved resolution would allow improved spectral patients as well as patients at different stages of
quality and less obscuration of the choline peak the menstrual cycle (Figs 1–5). All of these pa-
by either the water peak or the fat peak. tients have undergone imaging or clinical fol-
low-up for a year, with none having developed
Evaluation of Normal any abnormality at the site of spectroscopy. All
and Lactating Breast Parenchyma areas where the voxel was placed and 1H MR
Normal glandular parenchyma of the breast does spectroscopy was performed had MR imaging
not consistently demonstrate a choline resonance characteristics of normal breast parenchyma with-
peak that can be detected at 1.5 T. At our institu- out evidence of suspect enhancement. We did
tion, we conducted a small prospective study in not detect any choline in normal breast tissue at
which we performed 1H MR spectroscopy in 27 1.5 T.
patients undergoing screening breast MR imaging
(10). We included both pre- and postmenopausal
S244 October 2007 RG f Volume 27 ● Special Issue

Figure 3. (a) Sagittal non-fat-suppressed T1-weighted MR image (6.4/3.1) of the left

breast obtained in a 53-year-old woman shows normal glandular parenchyma. The patient
had undergone lumpectomy for cancer in the right breast that was discovered at screening
breast MR imaging performed owing to the patient’s high risk. (b) Magnified spectrum il-
lustrates a high lipid (Lip) peak, but no choline (Cho) resonance peak is observed at a fre-
quency of 3.2 ppm. Lac ⫽ lactate.

Figure 4. (a) Sagittal non-fat-suppressed T1-weighted MR image (6.4/3.1) of the left

breast obtained in a 56-year-old woman shows normal glandular parenchyma and little fat.
The patient had undergone lumpectomy for atypical ductal hyperplasia that was discovered
at screening breast MR imaging performed owing to the patient’s high risk. (b) Magnified
spectrum illustrates a high lipid (Lip) peak, but no choline (Cho) resonance peak is observed
at a frequency of 3.2 ppm. Lac ⫽ lactate.

A group in Australia studied 43 asymptomatic Differentiating Benign

volunteers, including three lactating mothers, at from Malignant Breast Lesions
1.5 T (4). A resonance in the choline spectral re- Studies have demonstrated that MR imaging can
gion (3.2 ppm) was observed in all three mothers, help detect otherwise occult breast cancers, and
but three false-positive resonances were also seen this modality is playing an increasingly important
in the remaining 40 (nonlactating) volunteers. role in the clinical setting, including a role in
Choline signal has also been documented in the screening high-risk women (13–15). Because of
lactating breast by other groups (11,12). Limited the lack of standardization of technique and inter-
data exist, but at higher field strengths, choline pretation, the specificity of MR imaging has been
signal can be detected in normal breast tissue, variable from center to center, but overall its
increasing the need for quantification of choline specificity has been relatively low, resulting in a
concentrations (3). The amount of choline de- considerable number of benign biopsies (16 –18).
tected in normal breast tissue at higher field Improving the positive predictive value (PPV) of
strengths has been less than in malignant lesions. MR imaging– based biopsy recommendations
would improve the acceptability and cost-effec-
tiveness of this imaging technique.
RG f Volume 27 ● Special Issue Bartella and Huang S245

Figure 5. (a) Sagittal non-fat-suppressed T1-weighted MR image (6.4/3.1) of the left

breast obtained in a 68-year-old woman shows normal glandular parenchyma and fat. The
patient had undergone contralateral lumpectomy for breast cancer that was discovered at
screening breast MR imaging performed owing to the patient’s high risk. (b) Magnified
spectrum illustrates a high lipid (Lip) peak, but no choline (Cho) resonance peak is observed
at a frequency of 3.2 ppm. Lac ⫽ lactate.

Findings in Selected In Vivo Single-Voxel 1H MR Spectroscopic Studies Performed on 1.5-T Imagers

True- True- False- False-

Malignant Benign Sensitivity Specificity Positive Negative Positive Negative PPV
Study* Lesions Lesions (%) (%) Findings Findings Findings Findings (%)
Roebuck et
al (19) 10 7 70 86 7 6 1 3 88
Kvistad et al
(11) 11 11 82 82 9 9 2 2 82
Cecil et al
(20) 23 15 83 87 19 13 2 4 90
Yeung et al
(21) 24 6 92 83 22 5 1 2 97
Jagannathan
et al
(12) 32 14 81 86 26 12 2 6 93
Tse et al
(22) 19 21 89 100 17 21 0 2 100
Huang et al
(7) 18 12 100 87 18 8 4 0 82
Bartella et al
(23) 31 26 100 88 31 23 3 0 91
Total 168 112 87† 87† 149 97 15 19 90†
*Numbers in parentheses indicate reference numbers.
†Average percentage.

1H MR spectroscopy has been suggested as an aimed at improving discrimination between be-

adjunct to breast MR imaging to improve the nign and malignant breast lesions have been con-
specificity of the latter technique. Prior studies ducted at several centers (1,6,7,11,19 –22,24,25).
performed on 1.5-T MR imagers have reported In a study that we conducted at our institution
sensitivities of 70%–100% and specificities of (23), breast 1H MR spectroscopy had a sensitivity
67%–100% for breast MR spectroscopy (Table). of 100% and a specificity of 88%, comparing
Multiple in vivo 1H MR spectroscopic studies
S246 October 2007 RG f Volume 27 ● Special Issue

Figure 6. Benign fibrosis and ductal hyperplasia (true-negative findings) in a 43-year-old

woman who presented with a new palpable mass in the right breast. (a) Sagittal contrast-
enhanced fat-suppressed T1-weighted MR image (6.4/3.1) demonstrates a 4.2-cm irregular
mass (arrow). A voxel was placed around the mass. (b) Magnified spectrum shows no posi-
tive choline (Cho) resonance peak, with only a noise level at a frequency of 3.2 ppm. Lac ⫽
lactate, Lip ⫽ lipid. MR-guided biopsy followed by surgical excision revealed benign fibrosis
and ductal hyperplasia. (Fig 6 reprinted, with permission, from reference 23.)

Figure 7. Mammographically detected, biopsy-proved invasive ductal carcinoma (true-

positive finding) of the left breast in a 52-year-old woman. (a) Sagittal fat-suppressed T1-
weighted MR image (6.4/3.1) obtained immediately after the intravenous injection of gado-
linium-based contrast material shows a 1.5-cm rim-enhancing mass. A voxel was placed
around the mass. (b) Magnified spectrum illustrates a positive choline (Cho) resonance peak
at a frequency of 3.2 ppm with an SNR greater than 2. Lac ⫽ lactate, Lip ⫽ lipid. (Fig 7 re-
printed, with permission, from reference 23.)

favorably with the results of prior reports that breast MR imaging, reducing the number of be-
made use of this technique. The use of 1H MR nign biopsies without compromising the diagnosis
spectroscopy as an adjunct to breast MR imaging of breast cancer (Fig 6).
would have significantly (P ⬍ .01) increased the All cancers in the study described in this article
PPV of biopsy from 35% to 82% and might have were identified at 1H MR spectroscopy; there
obviated biopsy in 57% of the 40 lesions with un- were no false-negative findings. A choline peak
known histologic features, with none of the can- was identified at 1H MR spectroscopy in a variety
cers being missed. These data suggest that 1H of cancer histologies, including 16 invasive can-
MR spectroscopy may be a useful supplement to cers (infiltrating ductal, infiltrating lobular, and
infiltrating mixed ductal and lobular carcinoma)
(Fig 7) and one ductal carcinoma in situ (DCIS).
RG f Volume 27 ● Special Issue Bartella and Huang S247

Figure 8. Fibroadenoma and fibroadenomatoid changes (false-positive findings) in a 43-

year-old woman with biopsy-proved DCIS. A suspect lesion was detected at MR imaging
performed to determine disease extent. (a) Sagittal contrast-enhanced fat-suppressed T1-
weighted MR image (6.4/3.1) shows regional clumped enhancement in the outer portion of
the left breast. (b) Magnified spectrum illustrates a choline (Cho) resonance peak with an
SNR greater than 2. Lac ⫽ lactate, Lip ⫽ lipid. Excision revealed fibroadenoma and fibroad-
enomatoid changes. (Fig 8 reprinted, with permission, from reference 26.)

Figure 9. Chronic inflammatory lesion with atypia (false-positive findings) in a 51-year-

old woman with a positive family history of breast cancer. The patient presented with a sus-
pect lesion that had been detected at screening breast MR imaging performed owing to the
patient’s high risk. (a) Sagittal contrast-enhanced fat-suppressed T1-weighted MR image
(6.4/3.1) of the left breast shows ductal clumped enhancement in the retroareolar region. A
voxel was placed around the area of enhancement. (b) Magnified spectrum illustrates a posi-
tive choline (Cho) resonance peak with an SNR greater than 2. Lac ⫽ lactate, Lip ⫽ lipid.
Excision of the lesion demonstrated an atypical chronic inflammatory lesion. (Fig 9 re-
printed, with permission, from reference 23.)

The latter lesion is of interest in light of prior re- changes (Fig 8), a chronic inflammatory lesion
ports suggesting that DCIS may not always dem- with atypia (Fig 9), and atypical ductal hyper-
onstrate a choline peak (19,24). Further study plasia with columnar cell alteration. A false-posi-
involving more DCIS lesions is essential. tive choline peak has previously been reported
This study included three false-positive find- with a fibroadenoma (11,21); to our knowledge,
ings: a fibroadenoma and fibroadenomatoid
S248 October 2007 RG f Volume 27 ● Special Issue

Figure 10. Palpable, mammographically detected, biopsy-proved invasive lobular carci-

noma (true-positive finding) in the left breast of a 56-year-old woman. (a) Sagittal fat-sup-
pressed T1-weighted MR image (6.4/3.1) obtained immediately after the intravenous injec-
tion of gadopentetate dimeglumine shows a 5-cm area of regional clumped enhancement in
the 12-o’clock axis. (b) Magnified spectrum illustrates a choline (Cho) resonance peak at a
frequency of 3.2 ppm with an SNR greater than 2. Lac ⫽ lactate, Lip ⫽ lipid. (Fig 10 re-
printed, with permission, from reference 26.)

Figure 11. Non–mass-enhancing lesion (true-negative finding) in a 38-year-old woman

with a BRCA1 gene. A suspect lesion was detected at screening MR imaging. (a) Sagittal
contrast-enhanced fat-suppressed T1-weighted MR image (6.4/3.1) obtained on day 11 of
the menstrual cycle shows focal clumped enhancement in the upper inner portion of the left
breast. (b) Magnified spectrum illustrates a high lipid (Lip) peak, but no choline (Cho) reso-
nance peak is observed at a frequency of 3.2 ppm. Lac ⫽ lactate.

however, no choline peak has been reported with atypia in these two lesions, excision would have
the other two lesions, although the number of been the standard of care. Further work is neces-
published series on single-voxel breast 1H MR sary to evaluate the prevalence and characteristics
spectroscopy is limited. In view of the presence of of false-positive findings at 1H MR spectroscopy.
Enhancing lesions at MR imaging that lead to
referral for biopsy are described as either mass
enhancing or non–mass enhancing. Non–mass
RG f Volume 27 ● Special Issue Bartella and Huang S249

Figure 12. Spectroscopy of non–mass-enhancing lesions.

(a) A suspect lesion was detected at screening in the left breast
of a 20-year-old woman with a positive family history of breast
cancer. Sagittal contrast-enhanced fat-suppressed T1-weighted
MR image (6.4/3.1) shows focal clumped enhancement in the
upper inner quadrant of the left breast. A voxel was placed
around the area of enhancement. Spectroscopy did not demon-
strate a choline resonance peak. MR-guided biopsy showed fi-
broadenomatoid change and breast parenchyma. (b) Benign
findings in a 43-year-old woman with a family history of breast
cancer (sister, age 41 years) who presented with nipple discharge
and breast pain. Mammography showed dense breasts without
suspect findings. Sagittal contrast-enhanced fat-suppressed T1-
weighted MR image (6.4/3.1) of the right breast demonstrates
unilateral regional enhancement at the 12-o’clock position. A
voxel was placed around the area of enhancement. MR spectros-
copy did not demonstrate a positive choline resonance peak.
MR-guided biopsy showed benign breast parenchyma. (c) DCIS
in a 57-year-old woman who presented with bloody nipple dis-
charge from the right breast. No malignant findings were seen at
mammography. Sagittal contrast-enhanced fat-suppressed T1-
weighted MR image (6.4/3.1) of the right breast demonstrates
segmental clumped enhancement of the entire lower outer quad-
rant. A voxel was placed around the area of enhancement. Spectroscopy demonstrated a positive choline resonance
peak at a frequency of 3.2 ppm with an SNR greater than 2. MR-guided biopsy and subsequent mastectomy revealed
extensive DCIS with a high nuclear grade. (d) Fibrocystic change and ductal hyperplasia in a 60-year-old woman
with a history of lumpectomy of the right breast for DCIS. Sagittal contrast-enhanced fat-suppressed T1-weighted
MR image (6.4/3.1) of the left breast shows ductal clumped enhancement in the retroareolar region. A voxel was
placed around the area of enhancement. No choline resonance peak was detected at spectroscopy. Excision revealed
fibrocystic change and ductal hyperplasia. (Fig 12 reprinted, with permission, from reference 26.)

enhancement, defined as “enhancement of an entities but may also occur in malignancies

area that is not a mass,” may involve different- (27,28). Biopsy is often necessary to differentiate
sized areas, with internal enhancement that is dis- benign lesions with non–mass enhancement from
crete from normal enhancing breast parenchyma cancer (Figs 10 –12). However, few data have
(26). Non–mass enhancement has been described
in benign hormonal changes and other benign
 S250 October 2007 RG f Volume 27 ● Special Issue

addressed the application of breast 1H MR spec- quantification method. A more recent pilot study
troscopy to lesions with non–mass enhancement. was performed on a 4-T system. In this study, 13
Our preliminary data show that the information patients with locally advanced cancer were evalu-
obtained at 1H MR spectroscopy may decrease ated (a) before receiving their first chemotherapy
the number of biopsy recommendations for be- dose, (b) 24 hours after the first dose, and (c) af-
nign lesions with non–mass enhancement. In our ter the fourth dose. 1H MR spectroscopy was able Teaching
study, use of 1H MR spectroscopy as a supple- to help detect a change in the choline concentra- Point
Teaching
ment to breast MR imaging would have signifi- tion from baseline within 24 hours of administra-
Point
cantly increased the PPV of biopsy for MR imag- tion of the first dose of neoadjuvant chemo-
ing– detected lesions with non–mass enhance- therapy. This change had a positive correlation
ment from 20% to 63% and would have obviated with the change in final lesion size, with a statisti-
biopsy in 68% of lesions (29). cal significance of P ⫽ .001.
This hypothesis has also been evaluated at These results are indeed revolutionary, since
higher field strengths; at 4 T, a retrospective, 1H MR spectroscopy would be able to help pre-

blinded-observer performance study of 55 indi- dict clinical response in patients undergoing neo-
viduals was conducted. Performing 1H MR spec- adjuvant chemotherapy within 24 hours of their
troscopy in addition to MR imaging improved receiving the first dose. These results suggest that
sensitivity and specificity for all four readers and the addition of 1H MR spectroscopy may offer a
also improved interobserver agreement (3). These substantial advantage over MR imaging alone in
preliminary results are very promising, although the prediction of response to neoadjuvant chemo-
again, larger studies are needed for further evalu- therapy (31) and may ultimately enhance patient
ation. survival. Larger studies are being designed to fur-
ther evaluate these preliminary data.
Predicting Response
to Neoadjuvant Chemotherapy Current Limitations
Neoadjuvant chemotherapy, also known as pre- of Breast 1H MR Spectroscopy
operative, induction, or primary chemotherapy, is Research concerning breast 1H MR spectroscopy
administered prior to a definitive surgical exci- is rapidly expanding, and more and more exciting
sion. It is typically administered to patients with data are being reported. Considerable progress
locally advanced disease or large primary tumors. has been made: This technique is now well toler-
These patients include those with stage 3 and ated by patients in the clinical setting, with acqui-
stage 4 disease with isolated ipsilateral supracla- sition times of approximately 10 minutes. At
vicular adenopathy but no distant metastases present, however, breast 1H MR spectroscopy—
(30). Downstaging of disease may be achieved although promising—is not ready for clinical use.
with neoadjuvant chemotherapy, allowing surgi- As mentioned earlier, the single-voxel tech-
Teaching
cal excision of inoperable lesions. Neoadjuvant nique, which is the most commonly used tech-
Point
chemotherapy may allow breast conservation in nique, allows only one lesion to be examined at a
patients who would otherwise require mastec- time. In addition, the lesion must be around 1
tomy. It can also help identify patients who are cm3 in size for the data to be meaningful. In
resistant to standard chemotherapy as indicated breast evaluation, we often have to perform bi-
by the lack of response of the primary tumor and opsy on much smaller lesions, and overcoming
may serve as a surrogate for assessing the re- this limitation would be extremely important.
sponse of micrometastases and ultimately en- Most of the time, more than one lesion is ques-
hance patient survival. tioned on an MR image, so the ability to evaluate
An early pilot study has shown that with a multiple lesions or even the whole breast is some-
1.5-T magnet, a change in the total choline con- thing that we certainly hope to achieve in the fu-
centration was observed after the completion of ture.
neoadjuvant treatment, a finding that was con- Patients with a hematoma or a metallic clip
Teaching
firmed with pathologic analysis (12). A small must be excluded, since inhomogeneities of the
Point
group of 14 patients were evaluated in this pilot magnetic field are produced that affect spectros-
study, and detection of choline was used as the copy, which must be performed in a very homo-
geneous magnetic field. Patient motion also af-
fects this technique, so that short acquisition
times are essential (Fig 13). A spectroscopist is
still needed because off-line data processing must
RG f Volume 27 ● Special Issue Bartella and Huang S251

Figure 13. Biopsy-proved invasive ductal carcinoma in the left breast of a 57-year-old
woman. (a) Sagittal contrast-enhanced fat-suppressed T1-weighted MR image (6.4/3.1)
obtained immediately after the intravenous injection of gadopentetate dimeglumine shows a
1.4-cm irregular enhancing mass. (b) Magnified spectrum illustrates no choline (Cho) reso-
nance peak at a frequency of 3.2 ppm. Lac ⫽ lactate, Lip ⫽ lipid. Patient motion during the
examination is the most likely reason for the false-negative result.

be performed (off-line processing software is References

manufacturer dependent, not standardized); a 1. Bolan PJ, Meisamy S, Baker EH, et al. In vivo
more automated method would make MR spec- quantification of choline compounds in the breast
troscopy similar to any other MR imaging se- with 1H MR spectroscopy. Magn Reson Med
2003;50:1134 –1143.
quence, allowing it to be performed by the MR 2. Negendank W. Studies of human tumors by MRS:
imaging technologist and making it much easier a review. NMR Biomed 1992;5:303–324.
to incorporate into the daily clinical routine. 3. Meisamy S, Bolan PJ, Baker EH, et al. Adding in
Several groups are currently looking at cases of vivo quantitative 1H MR spectroscopy to improve
DCIS, but data on the ability of 1H MR spectros- diagnostic accuracy of breast MR imaging: pre-
liminary results of observer performance study at
copy to help detect these lesions as well as the 4.0 T. Radiology 2005;236:465– 475.
atypical lesions are still limited. Although all types 4. Stanwell P, Gluch L, Clark D, et al. Specificity of
of invasive cancers have been detected with 1H choline metabolites for in vivo diagnosis of breast
MR spectroscopy, larger studies are still needed cancer using 1H MRS at 1.5 T. Eur Radiol 2005;
to further evaluate these preliminary data (32). 15:1037–1043.
5. Hu J, Vartanian SA, Xuan Y, Latif Z, Soulen RL.
An improved 1H magnetic resonance spectro-
Conclusions scopic imaging technique for the human breast:
1H preliminary results. Magn Reson Imaging 2005;
MR spectroscopy appears to have a bright fu- 23:571–576.
ture in the field of breast imaging. Its role in dif- 6. Jacobs MA, Barker PB, Bottomley PA, Bhujwalla
ferentiating benign from malignant lesions and in Z, Bluemke DA. Proton magnetic resonance spec-
improving the specificity of breast MR imaging troscopic imaging of human breast cancer: a pre-
may result in fewer breast biopsies. Moreover, in liminary study. J Magn Reson Imaging 2004;19:
68 –75.
the setting of monitoring response to neoadjuvant 7. Huang W, Fisher PR, Dulaimy K, Tudorica LA,
chemotherapy, the results to date have been ex- O’Hea B, Button TM. Detection of breast malig-
tremely promising. Overcoming the current limi- nancy: diagnostic MR protocol for improved
tations of 1H MR spectroscopy is primarily a specificity. Radiology 2004;232:585–591.
technical challenge. We hope that interest in this 8. Joe BN, Chen VY, Salibi N, Fuangtharntip P,
Hildebolt CF, Bae KT. Evaluation of 1H-mag-
technique continues to grow, thereby encourag- netic resonance spectroscopy of breast cancer pre-
ing technologic advances and participation in and postgadolinium administration. Invest Radiol
clinical trials around the world. Large multicenter 2005;40:405– 411.
trials are increasingly needed and represent a vital
step in the establishment of this technique in the
clinical setting.
S252 October 2007 RG f Volume 27 ● Special Issue

9. Katz-Brull R, Lavin PT, Lenkinski RE. Clinical 22. Tse GM, Cheung HS, Pang LM, et al. Character-
utility of proton magnetic resonance spectroscopy ization of lesions of the breast with proton MR
in characterizing breast lesions. J Natl Cancer Inst spectroscopy: comparison of carcinomas, benign
2002;94:1197–1203. lesions, and phyllodes tumors. AJR Am J Roent-
10. Bartella L, Thakur S, Morris E, Liberman L, genol 2003;181:1267–1272.
Huang W, Dershaw D. Proton magnetic reso- 23. Bartella L, Morris EA, Dershaw DD, et al. Proton
nance spectroscopy (MRS) of the breast: does MR spectroscopy with choline peak as malignancy
normal breast parenchyma give a false positive marker improves positive predictive value for
choline peak? (abstr). In: Radiological Society of breast cancer diagnosis: preliminary study. Radiol-
North America scientific assembly and annual ogy 2006;239:686 – 692.
meeting program. Oak Brook, Ill: Radiological 24. Yeung DK, Yang WT, Tse GM. Breast cancer: in
Society of North America, 2005; 178. vivo proton MR spectroscopy in the characteriza-
11. Kvistad KA, Bakken IJ, Gribbestad IS, et al. Char- tion of histopathologic subtypes and preliminary
acterization of neoplastic and normal human observations in axillary node metastases. Radiol-
breast tissues with in vivo (1)H MR spectroscopy. ogy 2002;225:190 –197.
J Magn Reson Imaging 1999;10:159 –164. 25. Gribbestad IS, Singstad TE, Nilsen G, et al. In
12. Jagannathan NR, Kumar M, Seenu V, et al. vivo 1H MRS of normal breast and breast tumors
Evaluation of total choline from in-vivo volume using a dedicated double breast coil. J Magn Re-
localized proton MR spectroscopy and its response son Imaging 1998;8:1191–1197.
to neoadjuvant chemotherapy in locally advanced 26. American College of Radiology. ACR BI-RADS:
breast cancer. Br J Cancer 2001;84:1016 –1022. magnetic resonance imaging. In: ACR Breast Im-
13. Liberman L. Breast cancer screening with MRI: aging Reporting & Data System, breast imaging
what are the data for patients at high risk? N Engl atlas. Reston, Va: American College of Radiology,
J Med 2004;351:497–500. 2003.
14. Robson ME, Offit K. Breast MRI for women with 27. Bartella L, Liberman L, Morris EA, Dershaw DD.
hereditary cancer risk. JAMA 2004;292:1368 – Nonpalpable mammographically occult invasive
1370. breast cancers detected by MRI. AJR Am J Roent-
15. Morris EA, Liberman L, Ballon DJ, et al. MRI of genol 2006;186:865– 870.
occult breast carcinoma in a high-risk population. 28. Schnall MD, Blume J, Bluemke DA, et al. Diag-
AJR Am J Roentgenol 2003;181:619 – 626. nostic architectural and dynamic features at breast
16. Boetes C, Stoutjesdijk M. MR imaging in screen- MR imaging: multicenter study. Radiology 2006;
ing women at increased risk for breast cancer. 238:42–53.
Magn Reson Imaging Clin N Am 2001;9:357– 29. Bartella L, Thakur SB, Morris EA, et al. Enhanc-
372, vii. ing nonmass lesions in the breast: evaluation with
17. Morris EA. Review of breast MRI: indications and proton (1H) MR spectroscopy. Radiology 2007;
limitations. Semin Roentgenol 2001;36:226 –237. 245(1) (in press).
18. Orel SG, Schnall MD. MR imaging of the breast 30. Kuerer HM, Hunt KK, Newman LA, Ross MI,
for the detection, diagnosis, and staging of breast Ames FC, Singletary SE. Neoadjuvant chemo-
cancer. Radiology 2001;220:13–30. therapy in women with invasive breast carcinoma:
19. Roebuck JR, Cecil KM, Schnall MD, Lenkinski conceptual basis and fundamental surgical issues.
RE. Human breast lesions: characterization with J Am Coll Surg 2000;190:350 –363.
proton MR spectroscopy. Radiology 1998;209: 31. Meisamy S, Bolan PJ, Baker EH, et al. Neoadju-
269 –275. vant chemotherapy of locally advanced breast can-
20. Cecil KM, Schnall MD, Siegelman ES, Lenkinski cer: predicting response with in vivo 1H MR spec-
RE. The evaluation of human breast lesions with troscopy—a pilot study at 4 T. Radiology 2004;
magnetic resonance imaging and proton magnetic 233:424 – 431.
resonance spectroscopy. Breast Cancer Res Treat 32. Bartella L, Liberman L, Thakur S, Morris E,
2001;68:45–54. Huang W, Dershaw D. Magnetic resonance spec-
21. Yeung DK, Cheung HS, Tse GM. Human breast troscopy (MRS) of the breast: does sensitivity vary
lesions: characterization with contrast-enhanced in with cancer histology? [abstr]. In: Radiological
vivo proton MR spectroscopy—initial results. Ra- Society of North America Scientific Assembly and
diology 2001;220:40 – 46. Annual Meeting Program. Oak Brook, Ill: Radio-
logical Society of North America, 2005; 177.
RG Volume 27 • Special Issue • October 2007 Bartella and Huang

Proton (1H) MR Spectroscopy of the Breast

Lia Bartella, MD and Wei Huang, PhD
RadioGraphics 2007; 27:S241–S252 ● Published online 10.1148/rg.27si075504 ● Content Codes:

Page S242
The diagnostic value of 1H MR spectroscopy in cancer is typically based on the detection of elevated
levels of choline compounds, choline being a marker of active tumor (2).

Page S250
1
In our study, use of H MR spectroscopy as a supplement to breast MR imaging would have
significantly increased the PPV of biopsy for MR imaging–detected lesions with non–mass
enhancement from 20% to 63% and would have obviated biopsy in 68% of lesions (29).

Page S250
1
H MR spectroscopy was able to help detect a change in the choline concentration from baseline
within 24 hours of administration of the first dose of neoadjuvant chemotherapy.

Page S250
The single-voxel technique, which is the most commonly used technique, allows only one lesion to be
3
examined at a time. In addition, the lesion must be around 1 cm in size for the data to be
meaningful.

Page S250
Patients with a hematoma or a metallic clip must be excluded, since inhomogeneities of the magnetic
field are produced that affect spectroscopy, which must be performed in a very homogeneous
magnetic field. Patient motion also affects this technique, so that short acquisition times are essential
(Fig 13).