| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Russell H. Morgan Department of Radiology and Radiological Science, Johns
Hopkins University School of Medicine and Johns Hopkins Medical Institutions,
601 N Caroline St., JHOC 3171C, Baltimore, MD 21287.
2 Department of Radiology, Johns Hopkins University School of Medicine,
Baltimore, MD.
3 Division of MR Research, Department of Radiology, Johns Hopkins University
School of Medicine, Baltimore, MD.
4 Department of Radiology, Health Science Informatics, Johns Hopkins University
School of Medicine, Baltimore, MD.
5 Division of Orthopaedics and Oncology, Department of Orthopaedic Surgery,
Johns Hopkins Medical Institutions, Baltimore, MD.
6 Department of Pathology, Johns Hopkins Medical Institutions, Baltimore,
MD.
7 MRI Division, Russell H. Morgan Department of Radiology and Radiological
Science, Johns Hopkins Medical Institutions, Baltimore, MD.
Received July 24, 2006;
accepted after revision December 29, 2006.
Partially supported by assistance of the Young Investigator Award from the
Society of Computed Body Tomography and Magnetic Resonance (SCBT-MR).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. At 3 T, 18 patients with musculoskeletal lesions (four histologically proven to be malignant, 14 proven benign histologically or at clinical follow-up) underwent 23 MR spectroscopy studies, 20 with a single-voxel technique and three with a multivoxel technique. Seventeen patients were imaged with a surface coil and six with a body coil. Choline signal (3.2 ppm) was measured in each voxel and expressed relative to background noise as signal-to-noise ratio (SNR). Choline SNRs of malignant tumors and benign lesions were compared.
RESULTS. Diagnostic spectra were obtained in 20 of 23 lesions. For malignant lesions (osteosarcoma with two MR spectroscopy sites, metastasis, grade 1 sarcoma), choline SNRs were 5.2 and 4.2 (performed with body coil) and 4.8 and 18.7 (performed with surface coil), respectively. For benign lesions (neurofibroma, two stress reactions, bone cyst, hemangioma, lipoma, Baker cyst), choline SNR was 6.3 (with surface coil), 5.5 (with surface coil), and not detected for five cases. Seven postoperative patients with myocutaneous flaps showed either the typical spectrum of muscle or negligible choline. Only a water peak existed in a bone cyst and a significant lipid peak in a lipoma. Choline SNRs were different for malignant and benign lesions (11.7 vs 2.3, p = 0.04, as performed with a surface coil).
CONCLUSION. At 3 T, both single-voxel and multivoxel MR spectroscopy are feasible. Proton MR spectroscopy is a potential noninvasive tool for characterizing lesion composition and malignant activity.
Keywords: bone • MRI • MR spectroscopy • MR technique • oncologic imaging • soft-tissue neoplasms
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Proton MR spectroscopy has been used extensively to investigate tumor metabolism in other organ systems but has been used in the characterization of musculoskeletal lesions in a limited fashion, with only three published studies to date [4–6], all performed at 1.5 T. Although MR spectroscopy is feasible at 1.5 T, sensitivity for metabolite detection and quantification are expected to be improved at 3 T because of the expected increased signal-to-noise ratios (SNRs) and resolution compared with lower-field-strength scanners [7], although other factors (such as field homogeneity and magnetic susceptibility effects) may be unfavorable at higher field strengths.
Therefore, the purpose of this article is to show the feasibility of performing proton MR spectroscopy at 3 T for the clinical evaluation of musculoskeletal lesions and to determine its usefulness for characterizing lesions on the basis of their metabolic constituents. In this feasibility study, we evaluated both single-voxel and multivoxel in vivo proton MR spectroscopy at 3 T for the evaluation of bone and soft-tissue lesions.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Recruitment of Subjects and Patient Population
Eighteen patients (mean age, 47.8 years; range, 13–71 years; 13 male
and five female) with bone (six patients) and soft-tissue abnormalities (12
patients) were prospectively enrolled in this study. Lesions were classified
as malignant after biopsy confirmation. All benign lesions had histologic
confirmation or stable clinical and imaging appearance over time. Correlation
was made to all available imaging and clinical records to classify lesions as
benign or malignant.
Imaging and MR Spectroscopy Analysis
The MR spectroscopy sequence was added to routine MRI sequences that these
patients underwent for initial evaluation or posttreatment evaluation of their
abnormalities. MR spectroscopy was performed before contrast administration to
avoid possible changes in spectra due to contrast material
[8].
MRI was performed at 3 T and included conventional axial and coronal T1-weighted images (spin-echo; TR/TE, 690/15; field of view, 20 mm; slice thickness, 6 mm) along with axial fat-suppressed T2-weighted images (spin-echo; 2,886/100; field of view, 18 mm; slice thickness, 6 mm) and coronal STIR images (inversion recovery; 2,462/100; inversion time, 200 milliseconds; field of view, 20 mm; slice thickness, 6 mm) of the body part in question. Unenhanced and contrast-enhanced axial T1-weighted images (gradient-recalled echo; 8.7/4.3; flip angle, 90°; field of view, 18 mm; slice thickness, 6 mm) were obtained after the IV administration of gadolinium gadopentetate dimeglumine (0.1 mmol/kg) followed by a saline flush.
The single-voxel MR spectroscopy technique was used in 20 studies in 16 patients, and the multivoxel technique was used when this software became available (three studies in two patients). The imaging coil that was used was adapted to the size and extent of the lesion. For single-voxel-MR spectroscopy, five lesions were evaluated using the body coil and 15 with the flexible surface coil (FLEX-M, Philips Medical Systems); for the multivoxel technique, one lesion was evaluated with a body coil and two with a flexible coil.
Single-voxel, water-suppressed spectra were acquired using a point-resolved spectroscopy (PRESS) sequence (2,000/144) with 256 accumulations; bandwidth, 2,000 Hz; and scanning time, approximately 8.5 minutes. A single flexible surface coil was used for signal reception. The spectral data were obtained from an 8-mL voxel (2 x 2 x 2 cm) localized to the lesion in question by a radiologist with at least 4 and up to 8 years of experience in musculoskeletal tumor imaging. For postsurgical cases, the voxel was localized to the surgical site in an area of increased T2 signal or suspicious masslike abnormality when present, or over the surgical bed if no significant abnormal T2 signal was seen. Frequency selective water-suppression and spatial outer-volume-suppression saturation pulses were used.
Multivoxel MR spectroscopy was performed before contrast administration in a single 10-mm-thick section using PRESS. The scanning parameters were 2,000/144; 18 x 18 matrix size; slice thickness, 1 cm; 18 x 18 cm field of view; total data acquisition time, approximately 12 minutes. Nominal voxel size was 1.0 mL, and the voxels were localized to the lesion in question or, for postoperative cases, to the surgical bed. The echo signal was digitized with 512 data points and a spectral width of 2,000 Hz. Before multivoxel MR spectroscopy was performed, high-order shimming was performed to optimize field homogeneity, and water suppression was optimized using automated routines provided by the manufacturer. Water-suppression was accomplished with three sequential chemical shift selective pulses with a bandwidth of 75 Hz, applied on-resonance with the water signal.
Spectroscopy data were analyzed using software developed at our institution on an ULTRASPARC 60 computer system (Sun Microsystems). Spectra were processed with zero-order phase correction based on the water peak, exponential line broadening of 3 Hz, zero-filling by a factor of 8, and applying a high-pass convolution filter to remove the residual water signal in the time domain. After setting the chemical shift of water to 4.7 ppm, spectroscopic images were created by numeric integration over the following chemical shift ranges: 3.14–3.34 ppm for choline, 0–1.45 ppm for lipids, and 4.2–5.2 ppm for water. For display, metabolic images were linearly interpolated by a factor of 8. After baseline correction using a cubic spline, the peak height of the signal in the choline frequency range in one voxel located completely in the lesion was quantified using a simplex curve-fitting routine and expressed as a ratio relative to the background noise level found between 8.0 and 10.0 ppm (when no signals are expected) in the same voxel to determine the SNR.
Spectroscopic Analysis
The spectra were analyzed by a physicist who recorded whether each spectrum
was of diagnostic quality. Failure of the spectroscopy examination was defined
as a spectrum containing only noise without any identifiable metabolite peaks,
insufficient water suppression, excessive lipid contamination, or insufficient
field homogeneity. In the latter case, with an inhomogeneous magnetic field,
the resulting spectral lines were too broad and discrete metabolic peaks could
not be detected with accuracy. Peaks typically present in each voxel included
choline, creatine, water, and lipids. Metabolite choline SNRs in malignant
lesions and in nonmalignant lesions were reported as mean values. The voxel
with the maximum choline SNR was used for analysis of data from the multivoxel
technique. Choline SNR values between malignant and nonmalignant lesions were
compared using a two-sample Wilcoxon's rank-sum (or the Mann-Whitney) test.
Given that the body coil typically is three to five times less sensitive than
the surface coil (depending on voxel location), comparisons were performed
separately according to the coil that was used. Statistical significance was
set at p <0.05.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The final diagnoses of the soft-tissue abnormalities were one low-grade myxofibroblastic myxoid sarcoma, one recurrent high-grade sarcoma in a postoperative patient 3 months after the MR study, one lipoma, one neurofibroma, one Baker cyst, and no evidence of recurrent or residual tumor at follow-up after resection of soft-tissue sarcomas in seven postoperative patients. For soft-tissue lesions, one case of presumed neurofibroma in a patient with neurofibromatosis was stable for 24 months. The latter case was diagnosed by typical MRI features of a fusiform mass isointense to muscle on T1weighted imaging, hyperintensity on T2-weighted imaging, and central enhancement with a target sign after contrast administration. The Baker cyst showed typical features on MRI (fluid signal intensity between the medial head of the gastrocnemius tendon and the semimembranosus tendon and no evidence of internal contrast enhancement), although a negative percutaneous biopsy was performed at another institution. The lipoma and the low-grade myxofibroblastic myxoid sarcoma were diagnosed after surgical excision. One postoperative MRI (original histology was a high-grade soft-tissue sarcoma) showed a biopsy-proven recurrent tumor 3 months after surgery.
For the remaining seven postoperative patients, stability was documented by imaging and clinical follow-up, and no evidence of recurrent or residual tumor was seen at 9–25 months of follow-up (one patient was followed up for 9 months, one for 14 months, two for 16 months, one for 19 months, one for 24 months, and one for 25 months). For these seven patients, the original histology included five low-grade and two high-grade sarcomas. Four of these seven patients had a myocutaneous flap reconstruction after resection of the tumor.
Diagnostic quality spectra were obtained for 20 of 23 studies. Three failed studies (as defined by the criteria listed in the Subjects and Methods section) occurred in one malignant case (multivoxel MR spectroscopy in a postoperative patient with recurrent tumor performed with a body coil) and two benign cases (single-voxel MR spectroscopy performed with the flexible coil and the body coil in postoperative patients with no recurrence).
Tables 1, 2, 3 describe the lesion characteristics and patient population. Malignant lesions showed a discrete signal peak at 3.2 ppm assigned to choline-containing metabolites with choline SNR measurements of 4.8 and 18.6 (performed with the flexible surface coil) and 4.2 and 5.2 (performed with the body coil). The mean choline SNR for malignant lesions was 11.7 for cases performed with the flexible coil and 4.7 for cases performed with the body coil. Figure 1, 1F, 1G, 1H, 1I shows the spectrum obtained from a malignant lesion; corresponding histology is shown in Figures 1F, 1G, 1H. The histology is shown to highlight the ambiguity in the percutaneous biopsy results.
|
|
|
|
|
|
|
|
Regarding the benign abnormalities, the choline SNRs for the unicameral bone cyst, lipoma, spinal hemangioma, posttraumatic stress reaction, and Baker cyst were negligible (flexible coil and body coil), and for the neurofibroma and stress fracture were 6.3 and 5.5 (flexible coil), respectively. For the postoperative cases, the choline SNRs were 2.9 and 17.1 (with the flexible coil) and 3.1 (with the body coil) in three patients who had resection with myocutaneous flaps; for the remaining postoperative cases, the choline SNR was negligible. The mean choline SNR for benign abnormalities was 2.3 for cases performed with the flexible surface coil and 1.6 for cases performed with the body coil.
In addition to choline, other metabolites were detected. For example, patients with myocutaneous flaps showed a positive choline peak and a creatine peak in the pattern of a typical metabolic spectrum for normal muscle (Fig. 2). The spectrum obtained from the patient with a unicameral bone cyst showed a large water peak without evidence of other metabolites (Fig. 3), confirming the cystic nature of the lesion that was inconclusive on unenhanced MRI. The spectrum obtained in the patient with a lipoma showed a large lipid peak (Fig. 4) despite the presence of internal ambiguous signal in the lipomatous mass that made the diagnosis of a simple lipoma uncertain on conventional MRI.
|
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Clinical spectroscopy can be obtained from various nuclei, including hydrogen, phosphorus, sodium, and carbon. However, proton (hydrogen) MR spectroscopy has the greatest SNR, has the best spatial resolution, and is more easily integrated with MRI in a single examination than MR spectroscopy with other nuclei. Compared with 1.5 T, SNR and spatial resolution are increased at 3 T, thereby enabling the determination of the metabolic composition of a lesion with greater certainty. Therefore, although metabolites such as choline can be detected at 1.5 T, it is expected that quantification of these metabolites will be more readily performed at 3 T. The benefits of 3 T for improved SNR in some cases may not be realized because of changes in relaxation times and increased magnetic susceptibility effects [10]. Although to our knowledge ours is the first report of proton MR spectroscopy in the assessment of musculoskeletal lesions at 3 T, there are limited reports of the feasibility of proton MR spectroscopy in other organ systems [11–17].
In addition to magnetic field strength, many factors affect metabolite signal, including the type of coil, the distance between the voxel of interest and the coil, the size of the lesion, and the tissues involved. In this series, we report results of spectroscopy performed with both the body coil and a surface coil. The body coil has three to five times lower sensitivity than a surface coil. With improved signal at 3-T compared with 1.5-T MRI, we believed that body coil imaging at 3 T should be explored. The body coil is more easily implemented in practice than a surface coil because of increased anatomic coverage and, in some cases, lack of tailored coils at 3 T. Disadvantages of body coil imaging include suppressing the "out-of-voxel" signal that may confound intravoxel results. Indeed, good-quality spectra were gathered from the body coil in four of six cases, echoing a recent report by Li et al. [13] in the assessment of hepatic lesions.
Proton spectroscopy can be performed in two ways: as a single-voxel technique or as a multivoxel technique. The advantage of multivoxel MR spectroscopy, particularly compared with single-voxel proton MR spectroscopy, is that it can simultaneously obtain spectra over a large field of view. This provides information regarding lesion boundaries and infiltration into the surrounding tissues as well as variations in a lesion, such as areas of necrosis. With the single-voxel technique, information from only one portion of a lesion may be analyzed. Multivoxel MR spectroscopy has been used to quantify lipid-to-water ratios in vertebral bone marrow for determining the fraction of marrow lipid as a measure of bone weakness [18] and has been used in only one study for the characterization of musculoskeletal lesions [6].
A potential limitation of both single-voxel and multivoxel MR spectroscopy is the complex and inhomogeneous environment created by the presence of different tissue types in the musculoskeletal system, including muscle, fat, vessels, and osseous trabeculae. This environment may result in magnetic field inhomogeneity and susceptibility artifacts that cause overlap of the different signals. For example, the wings (or possibly gradient-induced modulation sidebands) resulting from much larger water and lipid peaks can potentially obscure adjacent metabolites of interest. Because chemical shift resolution (measured in Hz) is doubled at 3 T compared with 1.5 T, spectral resolution is expected to be improved, although this may be partially offset by decreased T2 relaxation times (which have been reported for brain metabolites compared with 1.5 T). SNR improvements at 3 T are also offset by T2 relaxation time losses, particularly when long TEs are used. The intermediate TE used here (TE, 144) was chosen as a compromise between good sensitivity and shorter TEs, with which water and lipid peaks may be too large [19].
The significance of the choline peak for the assessment of malignancy in different organ systems has already been established. For musculoskeletal lesions, Wang et al. [5] showed that choline could be reliably detected in large malignant bone and soft-tissue tumors using single-voxel MR spectroscopy at 1.5 T. Fayad et al. [6] showed that choline could be detected in malignant skeletal sarcomas using the multivoxel technique at 1.5 T. Choline is a precursor of acetylcholine and a component of the phospholipid metabolism of cell membranes. The choline peak visible at 3.2 ppm contains contributions from glycerophosphocholine, phosphocholine, and choline, all compounds that are involved in the synthesis and degradation of cell membranes. Their concentration may be affected in disorders that influence membrane turnover. Therefore, increased choline likely reflects increased membrane synthesis or an increased number of cells, both conditions that are seen in malignant tumors. The 3-T measurement of choline in low-grade and high-grade sarcomas and metastases in our study lends further credence to the notion that MR spectroscopy may be a reliable means of differentiating malignant from benign musculoskeletal tumors and may in fact occasionally contribute a layer of certainty to ambiguous percutaneous biopsy results (Fig. 1, 1F, 1G, 1H, 1I). No detectable choline peak was observed in most benign abnormalities, and overall a significant difference was seen in the choline levels between benign and malignant lesions. However, two benign abnormalities (neurofibroma and stress fracture) showed discrete choline peaks. This is not an unexpected finding given that benign abnormalities may be metabolically active.
Finally, unique spectral patterns were observed in several situations. First, in all postsurgical patients in whom a choline peak was detected, the typical spectrum of muscle was shown (a discrete choline peak and higher creatine peak), and all of these patients had myocutaneous flaps. Choline and creatine are present in metabolically active normal muscle, and because no evidence of recurrent tumor was seen in these patients, we concluded that this spectral pattern is the normal postoperative spectral appearance. The second spectral pattern we observed was that of negligible choline. The variation in choline content observed in these postoperative patients can be explained by voxel placement because the voxel was localized to the most suspicious area on conventional MRI (areas showing increased T2 signal presumably due to water content without residual or recurrent tumor) and, in the absence of such areas, was placed over the surgical bed in the region of the flaps when present. In addition, for the unicameral bone cyst and lipoma (Figs. 3, and 4), specific characteristics were shown on MR spectroscopy, whereas their conventional MRI results were atypical or inconclusive. It is expected that other metabolites may be detected in certain types of lesions; for example, according to a study by Oya et al. [4] in which 1.5 T was used, an unassigned metabolite at 2.0–2.1 ppm, possibly due to N-acetyl aspartate, could be detected in tumors of neurogenic origin.
Our study had some limitations. Because this was a feasibility study, there were several limitations in our technique and results. First, the choice of the MR coil was not uniform for all the cases. Performing MR spectroscopy with the body coil is simpler and less cumbersome than using the small flexible coil, and we were often unable to use this latter coil for large lesions or for certain body parts. Second, SNR was used as a relative measure of metabolite concentration. SNR comparisons between patients are limited because these measures are affected by many factors, including type of coil, size of the lesion, and distance between the lesion and the coil. Future work should include the performance of MR spectroscopy without water suppression and calculation of metabolite–water ratios as a simple method of quantification, or the use of other quantitation strategies should be explored.
Third, our sample size was small, although statistically significant results were obtained for some comparisons. Fourth, all malignant cases in this series had histologic confirmation, but several benign cases did not have a tissue diagnosis so we used clinical and imaging follow-up as confirmation of benign nature. It must be expected that in cases in which a lesion is thought to be benign, a histologic diagnosis is often not obtained. Finally, we treated repeated studies on the same patients (18 patients underwent 23 MR spectroscopy examinations) as independent observations. In two of these patients, the repeated MR spectroscopy studies were performed in the same region of interest and had the same result, whereas in other patients, MR spectroscopy was performed in different regions of interest. Therefore, although some of the repeated measures may be correlated, these likely do not affect the overall results.
In conclusion, this article reports the feasibility of performing MR spectroscopy at 3 T, its potential use in characterizing musculoskeletal lesions, and the methodology that should be successful in future studies for determining accurate sensitivity and specificity of MR spectroscopy for distinguishing malignant from benign musculoskeletal lesions. Further evaluation must be performed for both preoperative and postoperative patients to determine the usefulness of MR spectroscopy for these patients. In addition, future directions may include a comparison of the performance of MR spectroscopy on a 1.5-T system compared with a 3-T system to show the potential usefulness of MR spectroscopy at lower field strengths.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. L. Sah, R. Sharma, H. Kandpal, A. Seith, S. Rastogi, S. Bandhu, and N. R. Jagannathan In Vivo Proton Spectroscopy of Giant Cell Tumor of the Bone Am. J. Roentgenol., February 1, 2008; 190(2): W133 - W139. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
