- Email: [email protected]
Proton magnetic resonance spectroscopy of musculoskeletal lesions at 3 T with metabolite quantification
Clinical Imaging 34 (2010) 47 – 52
Proton magnetic resonance spectroscopy of musculoskeletal lesions at 3 T with metabolite quantification☆ Chan Wha Lee a,b , Joo-Hyuk Lee a,⁎, Dae Hong Kim a , Hye Sook Min a , Byung-Kiu Park a , Hwan Sung Cho a , Hyun Guy Kang a , Jin-Suck Suh b , Shigeru Ehara c a
Research Institute and Hospital, National Cancer Center, Goyang-si, Gyeonggi-do 410-769, South Korea b Department of Medicine, Yonsei University Graduate School, Seoul 120-752, South Korea c Department of Radiology, School of Medicine, Iwate Medical University, Morioka 020-8505, Japan Received 20 January 2009; accepted 10 March 2009
Abstract Purpose: To evaluate whether proton MR spectroscopy (MRS) at 3 T with metabolite quantification is helpful for characterizing musculoskeletal lesions and to reveal whether the concentration of choline is correlated with the pathologic degree of malignancy. Material and methods: Three-tesla MR images and proton MRS data from 27 consecutive patients with surgically proven musculoskeletal lesions were retrospectively analyzed. We analyzed the presence of choline peaks of malignant tumors according to the degree of malignancies, and we compared them with those of benign lesions. The concentrations of choline calculated by means of the linear combination of model spectra were also compared with respect to the degree of malignancy in each group. Results: The proton MRS had an overall sensitivity of 68.4%, specificity of 87.5%, positive predictive value of 92.3%, and negative predictive value of 53.8% for the detection of choline compounds. The positive detection rate for choline compounds in the primary malignancy group (53.8%) was lower than that of the metastasis group (100%). All false-negative results were shown in the Grade 1 primary malignancy group. There was no difference in the concentration of choline compounds with respect to the pathologic degree of differentiation. Conclusion: MR spectroscopy at 3 T with metabolite quantification is a helpful method that shows high specificity (87.5%) in characterizing musculoskeletal lesions, even though its sensitivity (68.4%) is relatively low. Grade 1 primary malignancies of bone and soft tissue tumor have a high potential for producing false-negative results. © 2010 Elsevier Inc. All rights reserved. Keywords: MR spectroscopy; Bone and soft tissue tumor; Bone and soft tissue malignancy; Pathologic grading; Metabolic quantification
1. Introduction Proton magnetic resonance (MR) spectroscopy provides information about the metabolism of malignant tumors in clinical settings. The effectiveness of proton MR spectroscopy is mainly based on the detection of elevated levels of choline compounds, which are a marker of active tumors [1]. Although proton MR spectroscopy has shown promise in the ☆
This work was supported in part by grants from the National Cancer Center, Korea (0710665-2). ⁎ Corresponding author. 111 Jungbalsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-do 410-769, South Korea. Tel.: +82 31 920 2636; fax: +82 31 920 2643. E-mail address: [email protected] (J.-H. Lee). 0899-7071/09/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.clinimag.2009.03.013
evaluation of malignant tumors, choline levels determined by biological aggressiveness may not be as greatly elevated in some cancers [2–4]. Proton MR spectroscopy has been utilized for the analysis of bone and soft tissue tumors [5–7]. With the advent of 3 T MRI, it is expected that MR spectra will be obtained with better signal-to-noise ratios (SNRs) with greater certainty. The feasibility of performing proton MR spectroscopy at 3 T for the evaluation of musculoskeletal lesions has been reported recently [8]. In the evaluation of proton MR spectroscopic data, lesion characterization may be performed to detect metabolites. However, the objective and reliable determination of metabolites and their concentrations is necessary [9]. Metabolite concentrations can be estimated by the linear
48
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
combination of model spectra (LCModel), which is a userindependent spectral fit program, requiring no subjective interactions [10]. In this study, we tried to (1) evaluate whether proton MR spectroscopy at 3 T with metabolite quantification is helpful for characterizing musculoskeletal lesions and (2) reveal whether the concentration of choline is correlated to the pathologic degree of malignancy.
2. Materials and methods 2.1. Subjects We retrospectively analyzed the MR images and proton MR spectroscopic data of 27 consecutive patients (mean age, 47.9 years; age range, 17–85 years) with musculoskeletal lesions, histologically confirmed by surgery between October 2006 and December 2007. There were 16 male (mean age, 49.3 years old; age range, 16–76 years) and 11 female (mean age, 45.7 years old; age range, 17–85 years) patients. Nineteen lesions were malignant tumors and eight were benign. The largest tumor dimension for malignant tumors ranged from 4.1 to 17.9 cm (mean size, 8.1 cm), and that of benign lesions ranged from 3.8 to 21.0 cm (mean size, 9.0 cm). All resected specimens were histologically examined by an experienced musculoskeletal pathologist (HSM). Tumor cell differentiation, necrosis, and mitosis were evaluated. We categorized malignant tumors as the primary malignancy and metastasis group. According to the aggressiveness of the tumors, the lesions of the primary malignancy group were graded as Grade 1, 2, or 3; those of the metastasis group were graded as well differentiated (WD), moderately differentiated (MD), or poorly differentiated (PD). The primary malignancy group consisted of eight men and five women, ranging in age from 17 to 85 years (mean age, 43.8 years old). The postoperative diagnosis of the primary malignancy group included osteosarcoma (n=3), Ewing sarcoma (n=2), malignant peripheral nerve sheath tumor (n=2), fibromyxoid sarcoma (n=1), myxoid liposarcoma (n=1), well-differentiated liposarcoma (n=1), chondrosarcoma (n=1), malignant hemangiopericytoma (n=1), and malignant fibrous histiocytoma (n=1). There were five Grade 1, four Grade 2, and four Grade 3 sarcomas. The metastasis group consisted of three men and three women, ranging in age from 54 to 76 years (mean age, 63 years old). The histopathological diagnoses of the primary site were non-small cell lung carcinoma (n=3), hepatocellular carcinoma (n=2), and follicular carcinoma of the thyroid gland (n=1). There were two WD, two MD, and two PD carcinomas. There were eight benign lesions. The postoperative diagnoses of benign lesions were lipoma (n=2), lipogranuloma (n=1), giant cell tumor of bone (n=1), elastofibroma (n=1), hemangioma (n=1), fibromatosis (n=1), and heterotopic ossification (n=1). Our institutional review board waived approval for this retrospective study. Informed
consent for the procedure was obtained from each patient after full explanation of the MR imaging technique before the examination. 2.2. MR imaging and proton MR spectroscopy MR imaging was performed with a 3-T whole-body unit (Achieva, Philips, Best, The Netherlands). Each patient was examined in the supine position. The body coil was used for the identification of tumors. An appropriate surface coil was selected for MR spectroscopy. Conventional MR imaging sequences for the diagnosis of musculoskeletal lesions were as follows: (1) axial and coronal T1-weighted turbo spinecho sequence (TSE) [repetition time (ms)/echo time (ms), 500/20; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 1; echo train length, 4; matrix size, 460–512×364–368], (2) axial and coronal or sagittal TSE sequence [repetition time (ms)/echo time (ms), 4500–5000/90; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 1–2; echo train length, 17–21; matrix size, 460–800×357–780], and (3) coronal fat saturated T2-weighted TSE sequence [repetition time (ms)/echo time (ms), 3000/90; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 2; echo train length, 10; matrix size, 364×280]. After the patient was given a bolus intravenous administration of 0.1 mmol/kg of gadopentate dimeglumine (Magnevist, Bayer Schering Pharma, Berlin, Germany) followed by a saline flush, contrast enhanced axial and coronal fat-saturated T1-weighted TSE images [repetition time (ms)/echo time (ms), 550/20; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 1; echo train length, 4; matrix size 360– 400×288–320] were acquired in the same position. Single-voxel, water-suppressed spectra were acquired using the point-resolved spectroscopic pulse sequence (TR/TE, 2000/144 ms) 15–20 min after the administration of contrast material. The size of voxels varied from 10×10×10 to 15×15×15 mm depending on the tumor size. The volume of interest (VOI) was carefully positioned within the enhancing solid lesion as seen on the contrast enhanced image. In three cases of nonenhancing benign lesions (two cases of lipoma and one case of heterotopic ossification), the VOI was positioned within the center of the solid lesion. We attempted to avoid inclusion of the adjacent normal bone and cortex when possible. Automated optimization of frequency and receiver gain adjustments, shimming, gradient tuning, and water suppression were used. Data were acquired at a spectral band width of 2000 Hz, and 128 signals were averaged to achieve an adequate SNR. 2.3. MR spectroscopy analysis All spectra were processed using the commercial MR software package provided by the manufacturer (Program of the Philips Package System). The spectra were analyzed
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
49
Fig. 1. MR spectrum of a 76-year-old man with a metastatic lesion in the left femur (primary lesion: hepatocellular carcinoma). (A) The voxel was positioned within the highly enhancing lesion of the proximal metaphysis of the left femur (upper row, rectangularly outlined area). A fitted spectrum from the scanner showed a positive choline peak at 3.2 ppm (lower row, arrow). (B) An LCModel fitted spectrum showed a positive choline peak at 3.2 ppm (arrow).
50
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
by an experienced physicist (DHK). To calculate metabolite concentrations from MR spectroscopic imaging data, MR spectroscopic raw data files were transferred to an off-line workstation for postprocessing with the LCModel (Provencher, 2001); this fitting algorithm applies an automatic linear combination of model spectra [10]. This userindependent program can estimate concentrations of even minor metabolites to high internal precision, and the fitting error is expressed as the standard deviation (SD). Metabolite information with an error below 20% of the SD was included in the final analysis. Choline, citrate, water, and lipid peaks were evaluated. A spectrum containing only noise without identifiable metabolic peaks, insufficient water suppression, excessive lipid contamination, or insufficient field homogeneity was considered to indicate failure of the spectroscopy examination, according to the report of Fayad et al. [8]. The concentrations of metabolites were estimated and expressed in international units allowing direct comparison between spectra acquired on the same system. We analyzed the spectral pattern of choline peaks of malignant tumors according to the degree of malignancies and compared them with those of benign lesions.
The concentrations of choline calculated by means of the LCModel program for the selected tumor voxels were also compared with respect to the degree of malignancy in each group using the Kruskal–Wallis test. Pb.05 was considered to indicate a statistically significant difference.
3. Results A positive choline peak was present in 14 (51.9%) of the 27 lesions. Thirteen (68.4%) of the 19 patients with malignant tumors showed a positive peak for choline compounds (Fig. 1). Seven (87.5%) of the eight benign lesions showed a negative peak for choline compounds (Fig. 2), whereas one patient with a giant cell tumor of the femur showed a positive peak for choline compounds. Proton MR spectroscopy had an overall sensitivity of 68.4%, specificity of 87.5%, positive predictive value of 92.8%, and negative predictive value of 53.8% for the detection of choline compounds. The positive detection rate for choline compounds in the primary malignancy group was lower than that of the metastasis group. In the primary malignancy group, seven
Fig. 2. MR spectrum of an 18-year-old woman with fibromatosis in the right gluteus maximus muscle. (A) The voxel was positioned within the highly enhancing lesion of the right gluteus maximus muscle (upper row, rectangularly outlined area). A fitted spectrum from the scanner showed a negative or equivocal choline peak at 3.2 ppm (arrow). (B) The LCModel-fitted spectrum showed a negative choline peak at 3.2 ppm (arrow).
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52 Table 1 Results of the choline peak Primary malignancy
Metastasis
Benign Grade 1 Grade 2 Grade 3 WD MD PD Choline (+) 1 Choline (−) 7 Total 8
– 5
4 – 13
3 1
2 –
2 – 6
2 –
14 13 27
(53.8%) of the 13 patients showed a positive peak for choline compounds; in contrast, all six of the patients in the metastasis group showed a positive peak. All false-negative results were shown in the primary malignancy group. In the primary malignancy group, choline was detected in all of four patients in Grade 2, and three of four patients in Grade 3, and none of the five patients in Grade 1. The pathologic diagnosis of Grade 1 lesions that showed false-negative results included well-differentiated liposarcoma, myxoid liposarcoma, fibromyxoid sarcoma, chondrosarcoma, and malignant hemangiopericytoma. In Grade 3 lesions, one case of osteoblastic osteosarcoma showed a false-negative result. In the metastasis group, choline was detected in all WD (n=2), all MD (n=2), and all PD lesions (n=2) (Table 1). The calculated choline concentrations in malignant tumors are given in Fig. 3. There was no statistically significant difference in choline concentration between Grade 2 and Grade 3 malignant tumors in the primary malignancy group. Additionally, we found no statistically significant differences in the choline concentrations with respect to the degree of differentiation in the metastasis group.
51
decreased T2 relaxation times. In addition, the lack of homogeneity in the magnetic field and susceptibility to artifacts that cause the overlap of different signals are also increased at 3 T. Fayad et al. [8] used the intermediate TE of 144 ms as a compromise between good sensitivity and shorter TEs, and these authors reported that proton MR spectroscopy was useful as a noninvasive tool for characterizing musculoskeletal lesions at 3 T. In our study using a single-voxel spectroscopy at 3 T with TE of 144 ms, choline compounds were detected in musculoskeletal malignancies with a sensitivity of 68.4%, specificity of 87.5%, positive predictive value of 92.8%, and negative predictive value of 53.8%. The positive detection rate for choline compounds was relatively low in our study in comparison to the report by Wang et al. [6]. A possible explanation for this discrepancy centers on the fact that all of the Grade 1 lesions in our primary malignancy group showed negative results for choline compounds, thereby lowering the positive detection rate. Wang et al [6] reported that one case of parosteal osteosarcoma, most of which was densely ossified, showed a false-negative result. These authors suggested that lower proton levels and susceptibility effects due to mineralization
4. Discussion Proton MR spectroscopy has been used as a supportive tool for conventional MR imaging in the detection of malignant lesions. Elevation of choline compounds has been identified in a variety of malignant tumors. The choline signal found at 3.2 ppm in the MR spectra is composed of glycerophosphocholine, phosphocholine, and choline. Choline is a precursor for the neurotransmitter acetylcholine and the membrane constituent phosphatidylcholine [11]. Elevation of the choline peak represents increased membrane biosynthesis and reflects cellular proliferation [1,12]. Wang et al. [6] reported that choline could be reliably detected in large malignant bone and soft tissue tumors by using a multiecho point-resolved spectroscopic protocol at 1.5 T that had a sensitivity of 95%, specificity of 82%, and accuracy of 89%. Fayad et al. [7] reported that choline could be detected in malignant skeletal sarcomas by using the multivoxel technique at 1.5 T. With the use of a 3-T MR unit, a high SNR and increasing spatial resolution can be achieved in MR spectroscopy. However, these advantages may be offset by
Fig. 3. The calculated choline concentrations. (A) In primary malignant tumors (G1: Grade 1; G2: Grade 2; G3: Grade 3). (B) In metastasis. There was no statistically significant difference in the concentrations of choline with respect to the degree of differentiation.
52
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
may account for the false-negative result. In our study, one false-negative case of Grade 3 lesion in the primary malignancy group involved an osteoblastic osteosarcoma, which contained a relatively large amount of osteoid and ossification. The choline peak can be detected in benign lesions that show hypercellularity, hypervascularity, and a large number of inflammatory cells [6,13–16]. Wang et al. [6] reported that benign lesions, such as perineuroma, giant cell tumor, and abscess, showed positive choline peaks. In our study, one giant cell tumor showed a positive peak for choline compounds. Several authors have investigated the relationship between the levels of metabolites and the degree of malignancies [4,17–19]. Leach et al. [19] revealed that the degree of elevation of choline compounds was related to the grade of the tumor, with higher levels being found in higher grade lesions. If MR spectroscopy is intended to be a reliable technique for the evaluation of tumors, however, objective quantification is essential. LCModel based on linear combination offers a good candidate for the quantification of metabolites [10]. This user-independent program can estimate concentrations of even minor metabolites to a high internal precision. Stadlbauer et al. [4] used LCModel in their study and reported that significantly lower choline levels were found in Grade 2 tumors than in Grade 3 tumors. Baik et al. [20] reported that the internal method using water as a reference could be used for accurately quantifying choline concentrations in breast cancer. We calculated metabolite concentrations using the LCModel in all of our lesions. Our results showed that there were no differences in the choline concentrations between Grade 2 and 3 lesions in the primary malignancy group, and no differences were observed between WD, MD, and PD lesions in the metastasis group. Because of the diversity of pathology in our study, there were limited numbers of specific disease entities. Because of this drawback, it is difficult to compare the choline concentrations of lesions with respect to the degree of malignancy in the same specific disease entity. Instead, we tried to observe the levels of the choline concentrations according to the degree of malignancy in both sarcomas and metastatic carcinomas. Further studies in a large series focused on a specific malignant tumor are necessary to overcome the weakness of this study.
5. Conclusion Our results demonstrate that MR spectroscopy at 3 T with metabolite quantification is a helpful method that shows high specificity (87.5%) in characterizing musculoskeletal lesions, even though its sensitivity (68.4%) is relatively low. Grade 1 primary malignancies of bone and soft tissue tumors have a high potential for falsenegative results.
References [1] Negendank W. Studies of human tumors by MRS: a review. NMR Biomed 1992;5:303–24. [2] Cecil KM, Schnall MD, Siegelman ES, Lenkinski RE. The evaluation of human breast lesions with magnetic resonance imaging and proton magnetic resonance spectroscopy. Breast Cancer Res Treat 2001;68: 45–54. [3] Yeung DK, Yang WT, Tse GM. Breast cancer: in vivo proton MR spectroscopy in the characterization of histopathologic subtypes and preliminary observations in axillary node metastasis. Radiology 2002; 225:190–7. [4] Stadlbauer A, Gruber S, Nimsky C, Fahlbusch R, Hammen T, Buslei R, Tomandl B, Moser E, Ganslandt O. Preoperative grading of gliomas by using metabolite quantification with high-spatial-resolution proton MR spectroscopic imaging. Radiology 2006;238:958–69. [5] Oya N, Aoki J, Shinozaki T, Watanabe H, Takagishi K, Endo K. Preliminary study of proton magnetic resonance spectroscopy in bone and soft tissue tumors: an assigned signal at 2.0–2.1 ppm may be a possible indicator of malignant neuroectodermal tumor. Radiat Med 2000;18:193–8. [6] Wang CK, Li CW, Hsieh TJ, Chien SH, Liu GC, Tsai KB. Characterization of bone and soft tissue tumors with in vivo 1H MR spectroscopy: initial results. Radiology 2004;232:599–605. [7] Fayad LM, Bluemke DA, McCarthy EF, Weber KL, Barker PB, Jacobs MA. Musculoskeletal tumors: use of proton MR spectroscopic imaging for characterization. J Magn Reson Imaging 2006;23:23–8. [8] Fayad LM, et al. Characterization of musculoskeletal lesions on 3-T proton MR spectroscopy. AJR Am J Roentgenol 2007;188:1513–2000. [9] Edelman RR, Hesselink J, Zlatkin M, et al, editors. Clinical Magnetic Resonance Imaging, 3rd ed., vol. 1. Philadelphia: Saunders, 2006. p. 477–81. [10] Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993;30:672–9. [11] Miller BL. A review of chemical issues in 1-H NMR spectroscopy: n-acetyl-L-aspartate, creatinine and choline. NMR Biomed 1991;4: 47–52. [12] Ruiz-Cabello J, Cohen JS. Phospholipid metabolites as indicators of cancer cell function. NMR Biomed 1992;5:226–33. [13] Rand SD, Prost R, Haughton V, et al. Accuracy of single voxel proton MR spectroscopy in distinguishing neoplastic from nonneoplastic brain lesions. AJNR Am J Neuroradiol 1997;18:1695–704. [14] Krouwer HG, Kim TA, Rand SD, et al. Single voxel proton MR spectroscopy of nonneoplastic brain lesion suggestive of neoplasm. AJNR Am J Neuroradiol 1998;19:1695–703. [15] Maheshwari SR, Mukherji SK, Neelon B, et al. The choline/creatinine ratio in five benign neoplasms. Comparison with squamous cell carcinoma by use of in vivo MR spectroscopy. AJNR Am J Neuroradiol 2000;21:1930–5. [16] Bitch A, Bruhn H, Vougiouka V, et al. Inflammatory CNS demyelination: histopathologic correlation with in vivo quantitative proton spectroscopy. AJNR Am J Neuroradiol 1999;20:1619–27. [17] Yang D, Korogi Y, Sugahara T, et al. Cerebral gliomas: prospective comparison of multivoxel 2D chemical-shift imaging proton MR spectroscopy, echoplanar perfusion and diffusion-weighted MRI. Neuroradiology 2002;44:656–66. [18] Vuori K, Kankaanranta L, Hakkinen AM, et al. Low-grade gliomas and focal cortical developmental malformations: differentiation with proton MR spectroscopy. Radiology 2004;230:703–8. [19] Leach MO, Verrill M, Glaholm J, et al. Measurements of human breast cancer using magnetic resonance spectroscopy: a review of clinical measurements and a report of localized 31P measurements of response to treatment. NMR Biomed 1998;11:314–40. [20] Baik HM, Sy MY, Yu H, et al. Quantification of choline-containing compounds in malignant breast tumors by 1H MR spectroscopy using water as an internal reference at 1.5 T. MAGMA 2006;19: 96–104.
Proton magnetic resonance spectroscopy of musculoskeletal lesions at 3 T with metabolite quantification☆ Chan Wha Lee a,b , Joo-Hyuk Lee a,⁎, Dae Hong Kim a , Hye Sook Min a , Byung-Kiu Park a , Hwan Sung Cho a , Hyun Guy Kang a , Jin-Suck Suh b , Shigeru Ehara c a
Research Institute and Hospital, National Cancer Center, Goyang-si, Gyeonggi-do 410-769, South Korea b Department of Medicine, Yonsei University Graduate School, Seoul 120-752, South Korea c Department of Radiology, School of Medicine, Iwate Medical University, Morioka 020-8505, Japan Received 20 January 2009; accepted 10 March 2009
Abstract Purpose: To evaluate whether proton MR spectroscopy (MRS) at 3 T with metabolite quantification is helpful for characterizing musculoskeletal lesions and to reveal whether the concentration of choline is correlated with the pathologic degree of malignancy. Material and methods: Three-tesla MR images and proton MRS data from 27 consecutive patients with surgically proven musculoskeletal lesions were retrospectively analyzed. We analyzed the presence of choline peaks of malignant tumors according to the degree of malignancies, and we compared them with those of benign lesions. The concentrations of choline calculated by means of the linear combination of model spectra were also compared with respect to the degree of malignancy in each group. Results: The proton MRS had an overall sensitivity of 68.4%, specificity of 87.5%, positive predictive value of 92.3%, and negative predictive value of 53.8% for the detection of choline compounds. The positive detection rate for choline compounds in the primary malignancy group (53.8%) was lower than that of the metastasis group (100%). All false-negative results were shown in the Grade 1 primary malignancy group. There was no difference in the concentration of choline compounds with respect to the pathologic degree of differentiation. Conclusion: MR spectroscopy at 3 T with metabolite quantification is a helpful method that shows high specificity (87.5%) in characterizing musculoskeletal lesions, even though its sensitivity (68.4%) is relatively low. Grade 1 primary malignancies of bone and soft tissue tumor have a high potential for producing false-negative results. © 2010 Elsevier Inc. All rights reserved. Keywords: MR spectroscopy; Bone and soft tissue tumor; Bone and soft tissue malignancy; Pathologic grading; Metabolic quantification
1. Introduction Proton magnetic resonance (MR) spectroscopy provides information about the metabolism of malignant tumors in clinical settings. The effectiveness of proton MR spectroscopy is mainly based on the detection of elevated levels of choline compounds, which are a marker of active tumors [1]. Although proton MR spectroscopy has shown promise in the ☆
This work was supported in part by grants from the National Cancer Center, Korea (0710665-2). ⁎ Corresponding author. 111 Jungbalsan-ro, Ilsandong-gu, Goyang-si, Gyeonggi-do 410-769, South Korea. Tel.: +82 31 920 2636; fax: +82 31 920 2643. E-mail address: [email protected] (J.-H. Lee). 0899-7071/09/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.clinimag.2009.03.013
evaluation of malignant tumors, choline levels determined by biological aggressiveness may not be as greatly elevated in some cancers [2–4]. Proton MR spectroscopy has been utilized for the analysis of bone and soft tissue tumors [5–7]. With the advent of 3 T MRI, it is expected that MR spectra will be obtained with better signal-to-noise ratios (SNRs) with greater certainty. The feasibility of performing proton MR spectroscopy at 3 T for the evaluation of musculoskeletal lesions has been reported recently [8]. In the evaluation of proton MR spectroscopic data, lesion characterization may be performed to detect metabolites. However, the objective and reliable determination of metabolites and their concentrations is necessary [9]. Metabolite concentrations can be estimated by the linear
48
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
combination of model spectra (LCModel), which is a userindependent spectral fit program, requiring no subjective interactions [10]. In this study, we tried to (1) evaluate whether proton MR spectroscopy at 3 T with metabolite quantification is helpful for characterizing musculoskeletal lesions and (2) reveal whether the concentration of choline is correlated to the pathologic degree of malignancy.
2. Materials and methods 2.1. Subjects We retrospectively analyzed the MR images and proton MR spectroscopic data of 27 consecutive patients (mean age, 47.9 years; age range, 17–85 years) with musculoskeletal lesions, histologically confirmed by surgery between October 2006 and December 2007. There were 16 male (mean age, 49.3 years old; age range, 16–76 years) and 11 female (mean age, 45.7 years old; age range, 17–85 years) patients. Nineteen lesions were malignant tumors and eight were benign. The largest tumor dimension for malignant tumors ranged from 4.1 to 17.9 cm (mean size, 8.1 cm), and that of benign lesions ranged from 3.8 to 21.0 cm (mean size, 9.0 cm). All resected specimens were histologically examined by an experienced musculoskeletal pathologist (HSM). Tumor cell differentiation, necrosis, and mitosis were evaluated. We categorized malignant tumors as the primary malignancy and metastasis group. According to the aggressiveness of the tumors, the lesions of the primary malignancy group were graded as Grade 1, 2, or 3; those of the metastasis group were graded as well differentiated (WD), moderately differentiated (MD), or poorly differentiated (PD). The primary malignancy group consisted of eight men and five women, ranging in age from 17 to 85 years (mean age, 43.8 years old). The postoperative diagnosis of the primary malignancy group included osteosarcoma (n=3), Ewing sarcoma (n=2), malignant peripheral nerve sheath tumor (n=2), fibromyxoid sarcoma (n=1), myxoid liposarcoma (n=1), well-differentiated liposarcoma (n=1), chondrosarcoma (n=1), malignant hemangiopericytoma (n=1), and malignant fibrous histiocytoma (n=1). There were five Grade 1, four Grade 2, and four Grade 3 sarcomas. The metastasis group consisted of three men and three women, ranging in age from 54 to 76 years (mean age, 63 years old). The histopathological diagnoses of the primary site were non-small cell lung carcinoma (n=3), hepatocellular carcinoma (n=2), and follicular carcinoma of the thyroid gland (n=1). There were two WD, two MD, and two PD carcinomas. There were eight benign lesions. The postoperative diagnoses of benign lesions were lipoma (n=2), lipogranuloma (n=1), giant cell tumor of bone (n=1), elastofibroma (n=1), hemangioma (n=1), fibromatosis (n=1), and heterotopic ossification (n=1). Our institutional review board waived approval for this retrospective study. Informed
consent for the procedure was obtained from each patient after full explanation of the MR imaging technique before the examination. 2.2. MR imaging and proton MR spectroscopy MR imaging was performed with a 3-T whole-body unit (Achieva, Philips, Best, The Netherlands). Each patient was examined in the supine position. The body coil was used for the identification of tumors. An appropriate surface coil was selected for MR spectroscopy. Conventional MR imaging sequences for the diagnosis of musculoskeletal lesions were as follows: (1) axial and coronal T1-weighted turbo spinecho sequence (TSE) [repetition time (ms)/echo time (ms), 500/20; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 1; echo train length, 4; matrix size, 460–512×364–368], (2) axial and coronal or sagittal TSE sequence [repetition time (ms)/echo time (ms), 4500–5000/90; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 1–2; echo train length, 17–21; matrix size, 460–800×357–780], and (3) coronal fat saturated T2-weighted TSE sequence [repetition time (ms)/echo time (ms), 3000/90; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 2; echo train length, 10; matrix size, 364×280]. After the patient was given a bolus intravenous administration of 0.1 mmol/kg of gadopentate dimeglumine (Magnevist, Bayer Schering Pharma, Berlin, Germany) followed by a saline flush, contrast enhanced axial and coronal fat-saturated T1-weighted TSE images [repetition time (ms)/echo time (ms), 550/20; slice thickness, 5 mm; slice spacing, 1 mm; field of view, 36–40 cm; number of excitation, 1; echo train length, 4; matrix size 360– 400×288–320] were acquired in the same position. Single-voxel, water-suppressed spectra were acquired using the point-resolved spectroscopic pulse sequence (TR/TE, 2000/144 ms) 15–20 min after the administration of contrast material. The size of voxels varied from 10×10×10 to 15×15×15 mm depending on the tumor size. The volume of interest (VOI) was carefully positioned within the enhancing solid lesion as seen on the contrast enhanced image. In three cases of nonenhancing benign lesions (two cases of lipoma and one case of heterotopic ossification), the VOI was positioned within the center of the solid lesion. We attempted to avoid inclusion of the adjacent normal bone and cortex when possible. Automated optimization of frequency and receiver gain adjustments, shimming, gradient tuning, and water suppression were used. Data were acquired at a spectral band width of 2000 Hz, and 128 signals were averaged to achieve an adequate SNR. 2.3. MR spectroscopy analysis All spectra were processed using the commercial MR software package provided by the manufacturer (Program of the Philips Package System). The spectra were analyzed
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
49
Fig. 1. MR spectrum of a 76-year-old man with a metastatic lesion in the left femur (primary lesion: hepatocellular carcinoma). (A) The voxel was positioned within the highly enhancing lesion of the proximal metaphysis of the left femur (upper row, rectangularly outlined area). A fitted spectrum from the scanner showed a positive choline peak at 3.2 ppm (lower row, arrow). (B) An LCModel fitted spectrum showed a positive choline peak at 3.2 ppm (arrow).
50
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
by an experienced physicist (DHK). To calculate metabolite concentrations from MR spectroscopic imaging data, MR spectroscopic raw data files were transferred to an off-line workstation for postprocessing with the LCModel (Provencher, 2001); this fitting algorithm applies an automatic linear combination of model spectra [10]. This userindependent program can estimate concentrations of even minor metabolites to high internal precision, and the fitting error is expressed as the standard deviation (SD). Metabolite information with an error below 20% of the SD was included in the final analysis. Choline, citrate, water, and lipid peaks were evaluated. A spectrum containing only noise without identifiable metabolic peaks, insufficient water suppression, excessive lipid contamination, or insufficient field homogeneity was considered to indicate failure of the spectroscopy examination, according to the report of Fayad et al. [8]. The concentrations of metabolites were estimated and expressed in international units allowing direct comparison between spectra acquired on the same system. We analyzed the spectral pattern of choline peaks of malignant tumors according to the degree of malignancies and compared them with those of benign lesions.
The concentrations of choline calculated by means of the LCModel program for the selected tumor voxels were also compared with respect to the degree of malignancy in each group using the Kruskal–Wallis test. Pb.05 was considered to indicate a statistically significant difference.
3. Results A positive choline peak was present in 14 (51.9%) of the 27 lesions. Thirteen (68.4%) of the 19 patients with malignant tumors showed a positive peak for choline compounds (Fig. 1). Seven (87.5%) of the eight benign lesions showed a negative peak for choline compounds (Fig. 2), whereas one patient with a giant cell tumor of the femur showed a positive peak for choline compounds. Proton MR spectroscopy had an overall sensitivity of 68.4%, specificity of 87.5%, positive predictive value of 92.8%, and negative predictive value of 53.8% for the detection of choline compounds. The positive detection rate for choline compounds in the primary malignancy group was lower than that of the metastasis group. In the primary malignancy group, seven
Fig. 2. MR spectrum of an 18-year-old woman with fibromatosis in the right gluteus maximus muscle. (A) The voxel was positioned within the highly enhancing lesion of the right gluteus maximus muscle (upper row, rectangularly outlined area). A fitted spectrum from the scanner showed a negative or equivocal choline peak at 3.2 ppm (arrow). (B) The LCModel-fitted spectrum showed a negative choline peak at 3.2 ppm (arrow).
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52 Table 1 Results of the choline peak Primary malignancy
Metastasis
Benign Grade 1 Grade 2 Grade 3 WD MD PD Choline (+) 1 Choline (−) 7 Total 8
– 5
4 – 13
3 1
2 –
2 – 6
2 –
14 13 27
(53.8%) of the 13 patients showed a positive peak for choline compounds; in contrast, all six of the patients in the metastasis group showed a positive peak. All false-negative results were shown in the primary malignancy group. In the primary malignancy group, choline was detected in all of four patients in Grade 2, and three of four patients in Grade 3, and none of the five patients in Grade 1. The pathologic diagnosis of Grade 1 lesions that showed false-negative results included well-differentiated liposarcoma, myxoid liposarcoma, fibromyxoid sarcoma, chondrosarcoma, and malignant hemangiopericytoma. In Grade 3 lesions, one case of osteoblastic osteosarcoma showed a false-negative result. In the metastasis group, choline was detected in all WD (n=2), all MD (n=2), and all PD lesions (n=2) (Table 1). The calculated choline concentrations in malignant tumors are given in Fig. 3. There was no statistically significant difference in choline concentration between Grade 2 and Grade 3 malignant tumors in the primary malignancy group. Additionally, we found no statistically significant differences in the choline concentrations with respect to the degree of differentiation in the metastasis group.
51
decreased T2 relaxation times. In addition, the lack of homogeneity in the magnetic field and susceptibility to artifacts that cause the overlap of different signals are also increased at 3 T. Fayad et al. [8] used the intermediate TE of 144 ms as a compromise between good sensitivity and shorter TEs, and these authors reported that proton MR spectroscopy was useful as a noninvasive tool for characterizing musculoskeletal lesions at 3 T. In our study using a single-voxel spectroscopy at 3 T with TE of 144 ms, choline compounds were detected in musculoskeletal malignancies with a sensitivity of 68.4%, specificity of 87.5%, positive predictive value of 92.8%, and negative predictive value of 53.8%. The positive detection rate for choline compounds was relatively low in our study in comparison to the report by Wang et al. [6]. A possible explanation for this discrepancy centers on the fact that all of the Grade 1 lesions in our primary malignancy group showed negative results for choline compounds, thereby lowering the positive detection rate. Wang et al [6] reported that one case of parosteal osteosarcoma, most of which was densely ossified, showed a false-negative result. These authors suggested that lower proton levels and susceptibility effects due to mineralization
4. Discussion Proton MR spectroscopy has been used as a supportive tool for conventional MR imaging in the detection of malignant lesions. Elevation of choline compounds has been identified in a variety of malignant tumors. The choline signal found at 3.2 ppm in the MR spectra is composed of glycerophosphocholine, phosphocholine, and choline. Choline is a precursor for the neurotransmitter acetylcholine and the membrane constituent phosphatidylcholine [11]. Elevation of the choline peak represents increased membrane biosynthesis and reflects cellular proliferation [1,12]. Wang et al. [6] reported that choline could be reliably detected in large malignant bone and soft tissue tumors by using a multiecho point-resolved spectroscopic protocol at 1.5 T that had a sensitivity of 95%, specificity of 82%, and accuracy of 89%. Fayad et al. [7] reported that choline could be detected in malignant skeletal sarcomas by using the multivoxel technique at 1.5 T. With the use of a 3-T MR unit, a high SNR and increasing spatial resolution can be achieved in MR spectroscopy. However, these advantages may be offset by
Fig. 3. The calculated choline concentrations. (A) In primary malignant tumors (G1: Grade 1; G2: Grade 2; G3: Grade 3). (B) In metastasis. There was no statistically significant difference in the concentrations of choline with respect to the degree of differentiation.
52
C.W. Lee et al. / Clinical Imaging 34 (2010) 47–52
may account for the false-negative result. In our study, one false-negative case of Grade 3 lesion in the primary malignancy group involved an osteoblastic osteosarcoma, which contained a relatively large amount of osteoid and ossification. The choline peak can be detected in benign lesions that show hypercellularity, hypervascularity, and a large number of inflammatory cells [6,13–16]. Wang et al. [6] reported that benign lesions, such as perineuroma, giant cell tumor, and abscess, showed positive choline peaks. In our study, one giant cell tumor showed a positive peak for choline compounds. Several authors have investigated the relationship between the levels of metabolites and the degree of malignancies [4,17–19]. Leach et al. [19] revealed that the degree of elevation of choline compounds was related to the grade of the tumor, with higher levels being found in higher grade lesions. If MR spectroscopy is intended to be a reliable technique for the evaluation of tumors, however, objective quantification is essential. LCModel based on linear combination offers a good candidate for the quantification of metabolites [10]. This user-independent program can estimate concentrations of even minor metabolites to a high internal precision. Stadlbauer et al. [4] used LCModel in their study and reported that significantly lower choline levels were found in Grade 2 tumors than in Grade 3 tumors. Baik et al. [20] reported that the internal method using water as a reference could be used for accurately quantifying choline concentrations in breast cancer. We calculated metabolite concentrations using the LCModel in all of our lesions. Our results showed that there were no differences in the choline concentrations between Grade 2 and 3 lesions in the primary malignancy group, and no differences were observed between WD, MD, and PD lesions in the metastasis group. Because of the diversity of pathology in our study, there were limited numbers of specific disease entities. Because of this drawback, it is difficult to compare the choline concentrations of lesions with respect to the degree of malignancy in the same specific disease entity. Instead, we tried to observe the levels of the choline concentrations according to the degree of malignancy in both sarcomas and metastatic carcinomas. Further studies in a large series focused on a specific malignant tumor are necessary to overcome the weakness of this study.
5. Conclusion Our results demonstrate that MR spectroscopy at 3 T with metabolite quantification is a helpful method that shows high specificity (87.5%) in characterizing musculoskeletal lesions, even though its sensitivity (68.4%) is relatively low. Grade 1 primary malignancies of bone and soft tissue tumors have a high potential for falsenegative results.
References [1] Negendank W. Studies of human tumors by MRS: a review. NMR Biomed 1992;5:303–24. [2] Cecil KM, Schnall MD, Siegelman ES, Lenkinski RE. The evaluation of human breast lesions with magnetic resonance imaging and proton magnetic resonance spectroscopy. Breast Cancer Res Treat 2001;68: 45–54. [3] Yeung DK, Yang WT, Tse GM. Breast cancer: in vivo proton MR spectroscopy in the characterization of histopathologic subtypes and preliminary observations in axillary node metastasis. Radiology 2002; 225:190–7. [4] Stadlbauer A, Gruber S, Nimsky C, Fahlbusch R, Hammen T, Buslei R, Tomandl B, Moser E, Ganslandt O. Preoperative grading of gliomas by using metabolite quantification with high-spatial-resolution proton MR spectroscopic imaging. Radiology 2006;238:958–69. [5] Oya N, Aoki J, Shinozaki T, Watanabe H, Takagishi K, Endo K. Preliminary study of proton magnetic resonance spectroscopy in bone and soft tissue tumors: an assigned signal at 2.0–2.1 ppm may be a possible indicator of malignant neuroectodermal tumor. Radiat Med 2000;18:193–8. [6] Wang CK, Li CW, Hsieh TJ, Chien SH, Liu GC, Tsai KB. Characterization of bone and soft tissue tumors with in vivo 1H MR spectroscopy: initial results. Radiology 2004;232:599–605. [7] Fayad LM, Bluemke DA, McCarthy EF, Weber KL, Barker PB, Jacobs MA. Musculoskeletal tumors: use of proton MR spectroscopic imaging for characterization. J Magn Reson Imaging 2006;23:23–8. [8] Fayad LM, et al. Characterization of musculoskeletal lesions on 3-T proton MR spectroscopy. AJR Am J Roentgenol 2007;188:1513–2000. [9] Edelman RR, Hesselink J, Zlatkin M, et al, editors. Clinical Magnetic Resonance Imaging, 3rd ed., vol. 1. Philadelphia: Saunders, 2006. p. 477–81. [10] Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993;30:672–9. [11] Miller BL. A review of chemical issues in 1-H NMR spectroscopy: n-acetyl-L-aspartate, creatinine and choline. NMR Biomed 1991;4: 47–52. [12] Ruiz-Cabello J, Cohen JS. Phospholipid metabolites as indicators of cancer cell function. NMR Biomed 1992;5:226–33. [13] Rand SD, Prost R, Haughton V, et al. Accuracy of single voxel proton MR spectroscopy in distinguishing neoplastic from nonneoplastic brain lesions. AJNR Am J Neuroradiol 1997;18:1695–704. [14] Krouwer HG, Kim TA, Rand SD, et al. Single voxel proton MR spectroscopy of nonneoplastic brain lesion suggestive of neoplasm. AJNR Am J Neuroradiol 1998;19:1695–703. [15] Maheshwari SR, Mukherji SK, Neelon B, et al. The choline/creatinine ratio in five benign neoplasms. Comparison with squamous cell carcinoma by use of in vivo MR spectroscopy. AJNR Am J Neuroradiol 2000;21:1930–5. [16] Bitch A, Bruhn H, Vougiouka V, et al. Inflammatory CNS demyelination: histopathologic correlation with in vivo quantitative proton spectroscopy. AJNR Am J Neuroradiol 1999;20:1619–27. [17] Yang D, Korogi Y, Sugahara T, et al. Cerebral gliomas: prospective comparison of multivoxel 2D chemical-shift imaging proton MR spectroscopy, echoplanar perfusion and diffusion-weighted MRI. Neuroradiology 2002;44:656–66. [18] Vuori K, Kankaanranta L, Hakkinen AM, et al. Low-grade gliomas and focal cortical developmental malformations: differentiation with proton MR spectroscopy. Radiology 2004;230:703–8. [19] Leach MO, Verrill M, Glaholm J, et al. Measurements of human breast cancer using magnetic resonance spectroscopy: a review of clinical measurements and a report of localized 31P measurements of response to treatment. NMR Biomed 1998;11:314–40. [20] Baik HM, Sy MY, Yu H, et al. Quantification of choline-containing compounds in malignant breast tumors by 1H MR spectroscopy using water as an internal reference at 1.5 T. MAGMA 2006;19: 96–104.