Investigations and research
Fundamental advantages at 7T allow
for fat composition quantification
I. Dimitrov
J. Ren
D. Sherry
C. Malloy
MR Clinical Science, Philips Healthcare, Cleveland OH, USA.
Advanced Imaging Research Center, UT Southwestern, Dallas TX, USA.
on, or adjacent to, double bonds at 1.5, or even
at 3T. Thus, the majority of MRS studies of
lipid content in human tissues have focused on
measuring fat fraction, calculated from the
ratio of total triglycerides to water, in nonwater suppressed spectra [9-11]. Alternatively,
1
H-decoupled 13C MRS has been developed as
a non-invasive alternative to biopsy for analysis
of adipose tissue composition [12-16]. The
chemical shift dispersion of 13C is a major
advantage in measuring lipid signals from the
various proton groups, but the need for a second
RF channel and proton decoupling limits its
clinical applicability. In order to achieve higher
chemical shift separation, two-dimensional
MRS has also been used at 3T for investigating
muscle lipid metabolism [17]. However, MRS
requires long acquisition times and may be of
limited clinical acceptance.
1
G
Figure 1. Fundamental advantages of
1
H magnetic resonance spectroscopy
at 7T. Chemical shift dispersion is
insufficient to adequately separate
the lipid peaks in this olive oil
phantom at 1.5T (thick black line).
In comparison, at 7T the peaks are
well-resolved (thin red line). Note
that the SNR at 7T is several times
higher than at 1.5T, as seen by
comparing the noise levels in the
3.0 - 3.5 ppm region. In addition, the
availability of third-order shims at
7T results in narrower line widths,
with ensuing additional increase in
the SNR. The chemical structure of
a representative triglyceride molecule
is shown, with all the protons
contributing MRS signals labeled
(A - J).
42
MEDICAMUNDI 53/1 2009
Western societies consume fat abundantly. The
Paleolithic motto of “eat plenty when food is
bountiful” is causing obesity rates of epidemic
proportions in affluent societies. Hand-in-hand
with excess calories come cardiovascular disease,
diabetes and cancer. It is also becoming
increasingly clear that what we eat matters as
much as how much we eat. Predisposition to
cancer [1, 2], type 2 diabetes [3-6] and heart
disease has been attributed to diets high in
saturated fats [7]. Current research has also
focused on osteoporosis being caused by “bone
obesity” [8]. Despite a considerable number of
studies, the precise causal relationship between
dietary lipids, adipose tissue composition and
the risk of disease remains elusive and difficult
to study, in part because of the traditional
requirement for invasive biopsy. The ability to
profile triglyceride composition non-invasively
using proton spectroscopy would allow for simple
integration of these tests with standard Magnetic
Resonance Imaging (MRI) exams. A systematic
analysis of fatty acid composition from 1H
magnetic resonance spectroscopy (MRS) in
humans has not been reported, likely because it
is difficult to resolve the signals from protons
In this article we report on a research method to
measure in vivo lipid composition by 1H MRS at
7T. We also report the research for a non-invasive
determination of iodine values of human fat.
The iodine value of tissue shows the oxidative
stability of the fat in the tissue and may prove
to be an important index in determining
predisposition to diseases involving oxidative
metabolism.
There are two fundamental advantages of
working at 7T that allow for fat composition
quantification:
• The most important advantage is the larger
chemical shift dispersion. Chemical shift
dispersion is ~B0. Given that in vivo signals
have line widths of 20 - 30 Hz, a separation of
at least 10 - 15 Hz is required in order to
distinguish reliably between two peaks. At 1.5T this means that peaks cannot reliably
be separated that are closer than 0.25 ppm. At 7T this translates to only 0.05 ppm! As an
example, in Figure 1, while signals D and E are
not distinguishable at 1.5T, and signal C
severely overlaps signal B, no such complications
are seen on the 7T spectrum.
• Higher signal-to-noise ratio (SNR). SNR ~B0.
For example, the 7T lipid protocol described
below takes less than one minute to acquire.
This reduces potential motion artifacts by
significantly reducing the total examination
time.
One of the biggest disadvantages at 7T – the
higher field inhomogeneities (ΔB ~ B0) – is
successfully mitigated on the Philips Achieva
7T platform with the availability of third-order
shimming hardware. Furthermore, some of the
high SNR can be traded for smaller voxels which
will improve the shim. Although the SNR is
~(voxel size)3, since the regional inhomogeneities
dominate the determination of the line width,
reducing the voxel dimensions will increase T2*
and this will partially compensate the SNR for
the decreased voxel size. It has been shown [18]
that the SNR falls only as ~(voxel size)2, due to
the increased T2*. Thus, decreasing the voxel size
by a factor of two will reduce the line widths by
two and the SNR only by four (not eight).
Methods
Single voxel 1H MR spectra of muscles in the
calf were obtained from healthy adults on a 7T
Achieva scanner (Philips Healthcare, Cleveland,
OH, USA) using a partial-volume quadrature
transmit/receive coil. The protocol was approved
by the local Institutional Review Board, informed
consent was obtained from the subjects and the
exam was well-tolerated by all subjects.
It is well-known that 1H MRS is a sensitive
reporter of the relative concentration of different
lipid components. Depending on the specific
fatty acids, deconvolution of the signal from a
mixture of fats is feasible using high-resolution
MRS. At low field strengths, and especially in
vivo, shimming, motion and imperfect water
suppression limit the chemical shift resolution of
triglycerides to mostly that of the methyl groups
at 0.9 ppm, methylene groups at 1.3 ppm and
olefinic resonances at 5.3 ppm. In comparison,
at 7T up to ten resonances can be distinguished,
designated with letters A - J in Figure 1. The
challenge therefore is to obtain maximal
information about fat composition from these
well-separated signals. Of the ten resonances,
six contribute equivalent information about the
total number of triglyceride molecules: the CH3
methyl protons (A); the CH2 methylene protons
α-(E) and β-(C) to the carbonyl; and the glycerol
backbone CH (I) and CH2 protons (G and H).
At the chemical shift resolution achievable at 7T,
there are four other informative resonances to
consider: 1) bulk CH2 methylene protons at 1.3 ppm (B); 2) allylic CH2 protons, α- to a
double bond, at 2.02 ppm (D); 3) diallylic (also
called bis-allylic) CH2 protons at 2.75 ppm (F);
and 4) olefinic, double bond -CH=CH- protons,
at 5.31 ppm (J), which partially overlap with the
glycerol CH methine proton at 5.21 ppm (I).
The 1H chemical shift of in vivo fat resonances
from bone marrow and subcutaneous tissue was
assigned such that the methyl signal was at
0.9 ppm. Resonance areas were determined by
fitting the spectra with Voigt shapes (variable
proportions of Lorentzian plus Gaussian) after
phasing and baseline correction. Peak areas for
each individual resonance were corrected with
its corresponding T1 and T2. After relaxation
corrections, the lipid composition was evaluated
as follows.
Although signals from a high-resolution 1H NMR
spectrum of triglyceride mixtures may be resolved
into most constituents, in vivo spectroscopy is
constrained by limited chemical shift resolution.
The problem in vivo can be simplified, without
losing essential information, by assuming that all
fatty acids contain either 0, 1 or 2 double bonds.
These three types of fatty acids account for
~97% - 98% of total fat. Biologically, trilinolenic
acid (18:3) is the only fatty acid that is excluded
in this simplification but it contributes only
~0.5% of the total triglycerides [19]. With this
assumption,
fsat + fmono + fpoly = 100%, (1)
where fsat, fmono and fpoly refer to the fraction of
fatty acids that are saturated, monounsaturated
and polyunsaturated, respectively. In the
following, letters A - J represent the areas of the
respective MRS signals. First, the signal unit
area per proton can be arbitrarily assigned as
U = E/6, since there are always six methylene
protons α- to COO (resonance E) in every
triglyceride molecule. There are alternative
ways of determining the unit proton area from
the spectra and one of them is utilized later in
determining the iodine value (IV). The fraction
that is polyunsaturated, fpoly, can be calculated
directly from the relative area of the resonance
of the six “bridging” diallylic protons
(resonance F):
fpoly = (F/6)/(E/6) = F/E
E Predisposition to cancer
and heart disease has
been attributed to diets
high in saturated fats.
E This article reports on
a method to measure in
vivo lipid composition
by 1H MRS at 7T.
(2)
Once fpoly is known, fmono can be evaluated from
the relative area of protons α- to a double bond
(resonance D) by:
fmono = (D/12)/(E/6) – fpoly = 0.5 * D/E – fpoly, (3)
where resonance D is divided by 12, since this
is the number of protons α- to a double bond in
MEDICAMUNDI 53/1 2009
43
every unsaturated triglyceride molecule. The fraction fpoly is subtracted, since protons α- to a double
bond in polyunsaturated triglycerides also contribute to the D resonance. The remaining unknown,
fsat, can be calculated from Equation 1.
E Deconvolution of the
signal from a mixture of
fats is feasible using
high-resolution MRS.
Finally, in cases where all ten resonances (A - J) are observed, the IV of lipid tissues can be computed
from the average molecular weight, as calculated from the spectra and from the total number of
double bonds [20]. The definition of IV is the mass of iodine (in grams) that will react with the
double bonds of 100 grams of fat:
of
IV = 100 * (MW I2) * ( number
double bonds )/(MWfat),
(4)
where MWI2 = 253.8 and MWfat are the molecular weights of an iodine molecule and of fat, respectively.
The average molecular weight of fat can be calculated from the spectra:
MWfat = [15.034*A + 14.026*(B+C+D+E+F)/2 + 173.1*Uav + 26.04*(J+I–Uav)/2]/Uav
Here 15.034 is the molecular weight of CH3,
14.026 of CH2, 173.1 of the glycerol backbone
plus the three carbonyl groups (C6H5O6) and
26.04 of CH=CH. In this calculation, the unit
average proton area is designated as:
Uav = [A/9 + C/6 + E/6 + (G + H)/4 + I]/5 (6)
Figure 2. In vivo 1H magnetic
resonance spectrum of human tibial
marrow at 7T. The insert shows the
position of the acquired voxel. In the
spectrum (thick black line), ten
resonances (A - J) provide a good fit
(blue lines) and the respective peak
areas are used in Equations 1-3 to
calculate a profile of the lipid
composition.
H
2
44
MEDICAMUNDI 52/3 2008
This averaging is done in order to minimize
variations in signal intensity due to unknown
underlying macromolecular peaks and possible
rolling baseline.
Results
As seen in Figure 2, typically ten lipid resonances
are observed in the 7T 1H spectra from
physiological fats, together with a small water
peak. The lipid resonances are well-resolved at
(5)
7T, except for the double bond protons which
partially overlap the methine proton of the
glycerol backbone (Δδ = 0.1 ppm). For lipid
composition quantification, the signal intensities
were corrected for the difference in T1 and T2.
Proton resonances at different structural positions
have quite different values of T1 and T2, with T1
in the range of 0.39 - 1.13 s and T2 in the range
of 40 - 66 ms. The study showed that for
quantification of fat composition, the T2
correction is much more important than the T1
correction: for a typical echo time (TE) = 11 ms,
the relative T2 correction among the resonances
D, E and F accounts for ~4%, but reaches 8%
at TE = 20 ms and is as large as 18% at
TE = 40 ms. This compares to a correction of
only about 1% from the T1 correction for the
same three resonances at a recovery time
(TR) = 2 s. The fractions of saturated,
monounsaturated and polyunsaturated fat
constituents were calculated as given by
Equation 1-3: The composition of bone marrow
and adipose fat were similar, with an average
27% saturated, 48% monounsaturated and
25% polyunsaturated fractions. These values are
in agreement with values obtained by thin-layer
chromatography [19].
The lipid IV index was also determined from the
relative peak areas, as given in Equation 4, to be
75. The average lipid molecular weight was
calculated to be 912. This is about 7% higher
than the value of 855, which is the expected
average lipid molecular weight measured by
thin layer chromatography [20]. This
overestimation is probably due to contamination
of the peak intensities with additional intensities
from macromolecular pools with chemical shifts
overlapping the lipid resonances. In addition,
the fitted data is seen to overestimate the bulk
CH2 protons signal for which a single
Lorentzian/Gaussian does not truly represent
its line shape. Also, the rest of the lipid signals
were fitted as singlets, whereas it is known that
they are multiplets. Future work will incorporate
this additional knowledge in fitting the in vivo
spectra.
3
For the in vivo studies, a typical single voxel
MRS of marrow and subcutaneous tissue only
takes ~1 minute with excellent SNR which allows
for rapid testing of instrumental and protocol
reproducibility. This study showed that under
repeated scans, the spectral variation in the same
voxel was typically less than 5%. This is
comparable to, or less than, the typical
post-processing errors caused by fitting imperfect
line shapes, which could be as much as 10% in
total.
Future work and conclusions
In addition to measuring fat composition in
subcutaneous fat and bone marrow, investigation
is underway to profile muscle fat composition.
The complexity of fitting in vivo human muscle
spectra increases dramatically due to the effective
“doubling” of the spectra, with signals from both
extramyocellular (EMCL) and intramyocellular
(IMCL) lipids [21]. Figure 3 shows the rich
informational content of 7T in vivo human
muscle spectroscopy. The additional information
contained in the spectrum may be invaluable in
evaluating diabetes progression, as the connection
between insulin resistance and intracellular
lipids is well-known [22]. What is not known
is if the composition of the intracellular lipids
matters and 7T MRS will be essential in
answering such fundamental metabolic questions.
Recently, others have shown that high-quality
1
H NMR spectroscopy of mouse adipose tissue
is feasible at 7T [23]. The analysis used to
quantify fat composition was similar to that
described here. Like this earlier work, we chose
to present the data as saturated,
monounsaturated or polyunsaturated fractions.
We believe this description is more likely to be
accepted by clinicians than alternative
classifications. This analysis also follows closely
the widely popular nutritional facts information
printed on food labels. Such similarity may
facilitate further acceptance and understanding
amongst the general population.
G
Figure 3. Complexity of in vivo human
muscle spectrum at 7T. There is
~0.2 ppm susceptibility shift between
extramyocellular (EMCL) and
intramyocellular (IMCL) lipid
components. Additional metabolites,
such as creatine, carnitine, taurine
and carnosine, can also be resolved.
In conclusion, this 1H MRS study at 7T shows
that it is possible to quantify the fat composition
of human subcutaneous and bone marrow in a
fast (~1 minute) and non-invasive manner.
This has the potential to facilitate longitudinal
monitoring of lipid changes in healthy and
diseased tissues in clinical practice K
References
[1]Simonsen N, van ’t Veer P, Strain J, Martin-Moreno J, Huttunen
J, Navajas J, et al. Adipose Tissue Omega-3 and Omega-6 Fatty Acid
Content and Breast Cancer in the EURAMIC Study. European
Community Multicenter Study on Antioxidants, Myocardial
Infarction, and Breast Cancer. Am J Epidemiol. 1998; 147(4):
342-352.
[2]Shannon J, King I, Moshofsky R, Lampe J, Gao D, Ray R, et al.
Erythrocyte Fatty Acids and Breast Cancer Risk: A Case-Control
Study in Shanghai, China. Am J Clin Nutr. 2007; 85(4):
1090-1097.
[3]Storlien L, Higgins J, Thomas T, Brown M, Wang H, Huang X,
et al. Diet Composition and Insulin Action in Animal Models. Br J
Nutr. 2000; 83(Suppl 1): S85-S90.
[4]Manco M, Mingrone G, Greco A, Capristo E, Gniuli D, De
Gaetano A, et al. Insulin Resistance Directly Correlates with
Increased Saturated Fatty Acids in Skeletal Muscle Triglycerides.
Metabolism. 2000; 49(2): 220-224.
MEDICAMUNDI 53/1 2009
45
[5]Storlien L, Jenkins A, Chisholm D, Pascoe W, Khouri S, Kraegen E.
Influence of Dietary Fat Composition on Development of Insulin
Resistance in Rats. Relationship to Muscle Triglyceride and Omega-3
Fatty Acids in Muscle Phospholipid. Diabetes. 1991; 40(2):
280-289.
[6]Meyer K, Kushi L, Jacobs D Jr, Folsom A. Dietary Fat and
Incidence of Type 2 Diabetes in Older Iowa Women. Diabetes Care.
2001; 24(9): 1528-1535.
[7]Hu F, Willett W. Optimal Diets for Prevention of Coronary Heart
Disease. JAMA 2002; 288(20): 2569-2578.
[8]Rosen C, Bouxsein M. Mechanisms of Disease: Is Osteoporosis the
Obesity of Bone? Nat Clin Pract Pheumatol. 2006; 2(1): 35-43.
[9]Szczepaniak L, Babcock E, Schick F, Dobbins R, Garg A, Burns D, et al. Measurement of Intracellular Triglyceride Stores by
1
H Spectroscopy: Validation In Vivo. Am J Physiol. 1999; 276(5 Pt
1): E977-E989.
[10]Liney G, Bernard C, Manton D, Turnbull L, Langton C. Age,
Gender, and Skeletal Variation in Bone Marrow Composition: A
Preliminary Study at 3.0 Tesla. J Magn Reson Imaging. 2007;
26(3): 787-793.
[11]Yeung D, Griffith J, Antonio G, Lee F, Woo J, Leung P.
Osteoporosis is Associated with Increased Marrow Fat Content and
Decreased Marrow Fat Unsaturation: A Proton MR Spectroscopy
Study. J Magn Reson Imaging. 2005; 22(2): 279-285.
[12]Beckmann N, Brocard J, Keller U, Seelig J. Relationship Between
the Degree of Unsaturation of Dietary Fatty Acids and Adipose
Tissue Fatty Acids Assessed by Natural-Abundance 13C Magnetic
Resonance Spectroscopy in Man. Magn Reson Med. 1992; 27(1):
97-106.
[13]Dimand R, Moonen C, Chu S, Bradbury E, Kurland G, Cox K.
Adipose Tissue Abnormalities in Cystic Fibrosis: Noninvasive
Determination of Mono- and Polyunsaturated Fatty Acids by
Carbon-13 Topical Magnetic Resonance Spectroscopy. Pediatr Res.
1988; 24(2): 243-246.
[14]Moonen C, Dimand R, Cox K. The Noninvasive Determination of
Linoleic Acid Content of Human Adipose Tissue by Natural
Abundance Carbon-13 Nuclear Magnetic Resonance. Magn Reson
Med. 1988; 6(2): 140-157.
[15]Thomas E, Frost G, Barnard M, Bryant D, Taylor-Robinson S,
Simbrunner J, et al. An In Vivo 13C Magnetic Resonance
Spectroscopic Study of the Relationship Between Diet and Adipose
Tissue Composition. Lipids. 1996; 31(2): 145-151.
[16]Hwang J, Bluml S, Leaf A, Ross B. In Vivo Characterization of
Fatty Acids in Human Adipose Tissue using Natural Abundance 1H
Decoupled 13C MRS at 1.5 T: Clinical Applications to Dietary
Therapy. NMR Biomed. 2003; 16(3): 160-167.
46
MEDICAMUNDI 53/1 2009
[17]Velan S, Durst C, Lemieux S, Raylman R, Sridhar R, Spencer R,
et al. Investigation of Muscle Lipid Metabolism by Localized Oneand Two-Dimensional MRS Techniques using a Clinical 3T MRI/
MRS Scanner. J Magn Reson Imaging. 2007; 25(1): 192-199.
[18]Li B, Regal J, Gonen O. SNR Versus Resolution in 3D 1H MRS of
the Human Brain at High Magnetic Fields. Magn Reson Med.
2001; 46(6): 1049-1053.
[19]Field C, Angel A, Clandinin M. Relationship of Diet to the Fatty
Acid Composition of Human Adipose Tissue Structural and Stored
Lipids. Am J Clin Nutr. 1985; 42(6): 1206-1220.
[20]Miyake Y, Yokomizo K, Matsuzaki N. Rapid Determination of
Iodine Value by 1H Nuclear Magnetic Resonance Spectroscopy. J Am
Oil Chem Soc. 1998; 75: 15-19.
[21]Boesch C, Machann J, Vermathen P, Schick F. Role of Proton MR
for the Study of Muscle Lipid Metabolism. NMR Biomed. 2006;
19(7): 968-988.
[22]Stein DT, Szczepaniak LS, Dobbins R, Malloy CR, McGarry JD.
Skeletal Muscle Triglyceride Stores are Increased in Insulin Resistance.
(Abstract). Diabetes. 1997; 46 (Suppl 1): 23A.
[23]Strobel K, van den Hoff J, Pietzsch J. Localized Proton Magnetic
Resonance Spectroscopy of Lipids in Adipose Tissue at High Spatial
Resolution in Mice In Vivo. J Lipid Res. 2008; 49(2): 473-480.