Isaac R. Francis, MD #{149}Thomas William Ensminger, MD, PhD
Malignant Spectroscopy Chemical
L. Chenevert, PhD #{149} Barry Suzette Walker-Andrews,
neoplasms
P
of the
section
in
that
was occupied by tumor: less than 50% (group A) or more than 50% (group B). In group B, all phosphomonoester/-adenosine
triphosphate
studies
ratios
Radiology
1991;
shown
MR
spectroscopy
of liver
AND
METHODS
Healthy volunteers and patients whose liver tumors were detected with other inaging techniques and proved by means of biopsy were eligible for our study. To minimize the effects of therapy on the P-31 spectrum, the following study entry criteria were established retrospectively: (a) Patient had recently undergone themotherapy or radiation therapy (within 4 weeks) and had not responded, (b) therapy had failed and patient had progressive liver disease as determined with computed tomography, or (c) patient had stable or mildly progressive disease (usually from metastatic neuroendocrine tu-
Liver neoplasms, diagnosis, Liver neoplasms, MR studies, 761.1214. Magnetic resonance (MR), chemical shift #{149}Magnetic resonance (MR), phosphorus studies #{149} Magnetic resonance (MR), spectroscopy 761.33
on
MATERIALS
terms:
761.32,
resonance has
tumors (15,16). To address this issue, as well as to determine the feasibility and application of MR spectroscopy of the liver in clinical practice, we undertook this study of image-localized P-31 MR spectroscopy of malignant human liver tumors by using the onedimensional chemical shift imaging (CS!) technique (21). The methods, results, and limitations of this study form the basis of the present report.
were significantly higher than normal (P < .001). Hepatocellular carcinomas and metastases from various primary neoplasms could not be differentiated on the basis of spectral characteristics and metabolite ratios. Limitations of one-dimensional surface coil CSI prevented separation of spectra of small tumors and tumors deep within the liver parenchyma from spectra of normal liver parenchyma. Index
spectroscopy
#{149}
180:341-344
mors). I
From
the Departments
of Radiology
(I.R.F., T.LC.)
and Internal
Medicine
and Pharmacology
the DeHenry Ford Hospital, Detroit (B.C.); LaRavoire, France (L.C.); and the DeRadiology and Nudear Medicine, Stanford University, Stanford, Calif (G.M.G.). Received February 23, 1990; revision requested April 17; final revision received and accepted April 8,1991. Supported in part by the Tumor Imaging Core of the University of Michigan (WE., S.W.A.), University partment of Radiology, partment of Diagnostic
Cancer Center, NIH ©RSNA, 1991
of Michigan,
P30 CA 46592-01-Al.
1500
E Medical
Address
Center
reprint
Dr, Ann
requests
Arbor,
to LR.F.
Collomb, MD
MS
1 MR
potential utility in the evaluation of a variety of malignant neoplasms (116), and more recently its role has been extended to indude the momtoring of tumor response to therapy (17-20). Many studies have shown that, compared with normal tissues, malignant tumors of varying histologic descriptions have elevated levels of phosphomonoesters (PMEs) and, hence, elevated PMF43.-adenosine triphosphate (ATP) ratios (1-16). However, there are still relatively few
(30 metastases
analyzed
magnetic
HosiiORUs-31
(MR)
from a variety of primary tumors and seven hepatocellular carcinomas) and seven healthy volunteers. Tumors were grouped according to the percentage
MD #{149}Laurance Gary M. Glazer,
Hepatic Tumors: P-3 with One-dimensional Shift Imaging’
To determine the clinical feasibility and applicability of phosphorus-31 magnetic resonance (MR) spectroscopy and to assess its potential for characterization of human hepatic tissue, one-dimensional chemical shift imaging (CSI) was performed 37 patients with various malignant hepatic
Gubin, RN
#{149}
Thirty-seven patients with malignant liver tumors met one of these criteria and were induded in the study, as were seven healthy volunteers who served as controls. The 37 patients induded 30 with metastases from various primary tumors and seven with primary hepatocellular caranoma. The primary tumors in the 30 patients with liver metastases induded cobrectal cancer (n = 16), gastrointestinal neuroendocrine tumors such as carcinoids and islet cell tumors (n = 8), and a variety of other tumors (n 6). Patients with himors (metastases and hepatocellular cardnomas) ranged in age from 19 to 81 years, with a mean of 54 years. The ages of the volunteers ranged from 24 to 38 years, with a mean of 28 years. The metastases were 1-18 cm in diameter, with a mean of =
7 cm;
the hepatomas
were
1-14
cm in di-
ameter, with a mean of 8.5 cm. Spectroscopy was performed at the time of initial discovery of the tumor in seven of 30 patients with metastases and six of seven patients with hepatocellular cardnoma. No attempt was made to modify the diet of the patients or the volunteers before MR spectroscopy was performed. All studies system
Signa
were performed on a I .5-T (GE Medical Systems, Mil-
waukee). The imaging technique consisted of Ti-weighted axiallocalizing imaging (repetition time [TRI, 500 msec; echo time [TEl,
20 msec)
to define
the location
of the
for surface coil placement. With the patient supine, an 8- or 14-cm surface coil (depending on the size and location of the tumor) was placed on the skin surface, either along the anterior abdominal wall or the right lateral thoracoabdominal wall, over the lesions best suited for obtaining a tumors
P-31
MR spectra.
larger Optimal firmed
ages
and/or
These
were
superficially
usually
located
the
tumors.
coil positioning was then consubsequent TI-weighted inthat displayed a tube of high-signalwith
intensity material around the coil circumference. In the volunteers, the coil was placed on the right lateral thoracoabdominal wall for sampling
of the right
lobe.
MI 48109-0030;
Abbreviations: ATP = adenosine triphosphate, CS! = chemical shift imaging, PME phosphomonoester, S/N = signal-to-noise TE = echo time, TR = repetition time.
=
ratio,
341
Surface
coil-localized
proton
shimming
PME/I-ATP
was performed in approximately one-half of the studies by adjusting the first-order gradients. Shimming could not be performed in all cases because of personnel and time constraints. A P-31 one-dimensional
chemical
lected
along
ing
the
image
was
of signals
averaged,
128 phase-encoding 256
complex
full spectral coding
data
width.
pulse
0.4-msec
steps
(1-cm
points;
and
A 0.5-msec
followed thus
ma-
sec2-kHz
and
delayed
to
for
resolution the price
resolution to apthis phase-encoding is adequate paid in the
when signal-to-
noise ratio (S/N) and phase distortion required for finer resolution over longer phase-encoding periods. The total duralion of the studies ranged from 60 to 90 minutes, with the spectroscopy component of the study requiring 21 minutes 25 seconds of acquisition time and an addi-
5-10 minutes
of setup
time.
The P-31 data were processed ponential filtering (10 Hz), zero
with exfiffing,
two-dimensional Fourier transform to a 128 x 1,024 matrix, and manual phase correction.
The
from
baseline
signal
phase
distortion
digitization
encoding
was
delay reduced
that
results
due to by
means
of
a correction routine that used a third-order polynomial fit of the baseline. All of
these
steps
were
dor-provided the 1280 Data
ments, data
performed
software Station;
Fremont,
NMR
by using
yen-
(GEN ver.9.0 for GE NMR Instru-
Calif) on a spectroscopy
station Nicolet Instruments,
1280 computer Fremont, Calif).
ally only one spectrum
(GE Usu-
per one-dimen-
sional CS! data set was analyzed in detail. This spectrum was selected to match the depth of interest in the lesion as measured on proton images. While spectra were quantified by means of several methods (peak heights vs manual and automated fit of peak areas), problems
with
greatest
peak
overlap
confidence
curve-fitting
caused Only
mated fit of lorentzian curves by using NMRI spectroscopy cuse, puter
to have
the
to the peaks data analysis
(New Methods Research, SyraNY) operating on a VAX 1 1t750 com(DEC,
Maynard,
Mass).
The
PME,
inorganic phosphate, phosphodiester, phosphocreatine contaminant, and ATP peaks were quantified such that sets of ratios could be generated. Usually
342
only
Radiology
#{149}
one
spectrum
per
mensional
=
standard
Mean
SD 0.18 0.70 1.90
7
0.35
12 18 2
0.99
5
1.39
deviation,
1.81 1.15
NA
Ratio
NS
Bonnferoni Correction
Significance NA p P
1.34 0.64
not applicable,
=
one-di-
CS! data
set was analyzed
in
detail. This spectrum was selected by using the known 1-cm section spacing of the one-dimensional CS! data and the depth from the body wall (usually easily defined on one-dimensional CS! data) to the desired region in the lesion as measured on the proton image. Patients were grouped on the basis of the estimated percentage of the selected one-dimensional CS! section that was occupied by tumor. This area was
approximated
by superimposing
template over plate displayed
the proton images. The ternan idealized stack of 1-cm
one-dimensional
NA NS
.05