Magnetic Resonance Imaging 33 (2015) 644–648

Contents lists available at ScienceDirect

Magnetic Resonance Imaging journal homepage: www.mrijournal.com

Influence of echo time in quantitative proton MR spectroscopy using LCModel Tetsuya Yamamoto a, Tomonori Isobe b,⁎, Hiroyoshi Akutsu a, Tomohiko Masumoto c, Hiroki Ando b, Eisuke Sato b, Kenta Takada b, Izumi Anno d, Akira Matsumura a a b c d

Department of Neurosurgery, Faculty of Medicine, University of Tsukuba Graduate School of Comprehensive Human Sciences, University of Tsukuba Department of Radiology, Faculty of Medicine, University of Tsukuba Department of Radiological Sciences, Ibaraki Prefectural University of Health Sciences

a r t i c l e

i n f o

Article history: Received 7 May 2014 Revised 2 December 2014 Accepted 18 January 2015 Keywords: LCModel MRS Quantitative analysis Echo time T2 relaxation

a b s t r a c t Objective: The objective of this study was to elucidate the influence on quantitative analysis using LCModel with the condition of echo time (TE) longer than the recommended values in the spectrum acquisition specifications. Methods: A 3 T magnetic resonance system was used to perform proton magnetic resonance spectroscopy. The participants were 5 healthy volunteers and 11 patients with glioma. Data were collected at TE of 72, 144 and 288 ms. LCModel was used to quantify several metabolites (N-acetylaspartate, creatine and phosphocreatine, and choline-containing compounds). The results were compared with quantitative values obtained by using the T2-corrected internal reference method. Results: In healthy volunteers, when TE was long, the quantitative values obtained using LCModel were up to 6.8-fold larger (p b 0.05) than those obtained using the T2-corrected internal reference method. The ratios of the quantitative values obtained by the two methods differed between metabolites (p b 0.05). In patients with glioma, the ratios of quantitative values obtained by the two methods tended to be larger at longer TE, similarly to the case of healthy volunteers, and large between-individual variation in the ratios was observed. Conclusions: In clinical practice, TE is sometimes set longer than the value recommended for LCModel. If TE is long, LCModel overestimates the quantitative value since it cannot compensate for signal attenuation, and this effect is different for each metabolite and condition. Therefore, if TE is longer than recommended, it is necessary to account for the possibly reduced reliability of quantitative values calculated using LCModel. © 2015 Elsevier Inc. All rights reserved.

1. Introduction Magnetic resonance spectroscopy (MRS) is a diagnostic tool that can be used to analyze a variety of disease states. MRS is applicable to many nuclides, but proton MRS (utilizing the chemical shift of 1H nuclei (proton), which are ubiquitous in biological systems) provides higher signal intensity compared with MRS of other nuclides. Consequently, proton MRS has found clinical application in differential diagnosis for determining whether brain tumors are malignant or benign [1–3]. However, MRS is not as widely adopted as MR imaging because MRS requires an apparatus with highly uniform magnetic field for acquiring spectral data, the analysis of the spectral data is complex, and the reproducibility of spectral analysis results is low. In recent years, these ⁎ Corresponding author. Graduate School of Comprehensive Human Sciences, University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki 305-8575, Japan. Tel.: +81 29 853 7834; fax: +81 29 853 7102. E-mail address: [email protected] (T. Isobe). http://dx.doi.org/10.1016/j.mri.2015.01.015 0730-725X/© 2015 Elsevier Inc. All rights reserved.

problems are gradually being solved with the adoption of 3 T MR systems and the dedicated proton MRS analysis software LCModel (LCModel Version 6.3; http://s-provencher.com/pages/lcmodel.shtml). LCModel performs automatic analysis of spectral data and is capable of reducing variation in the analysis results caused by factors such as overlapping spectral peaks and distortion of the spectral baseline due to macromolecules. Therefore, LCModel is expected to contribute toward the standardization of analysis results [4,5]. To standardize the analysis results obtained using LCModel, spectral data must be collected under the recommended conditions. The recommended conditions for LCModel for minimizing the influence of relaxation time are repetition time (TR) of 6000 ms or greater and echo time (TE) of 30 ms or less [4,5]. However, there are cases in clinical practice where TE between 144 and 288 ms (which is longer than recommended) is used in order to differentiate between lactate and lipids or to ensure the reproducibility of acquired spectral data [6–10]. It is unclear how the analysis results are affected by using TE that is longer than recommended for LCModel, and quantitative values

T. Yamamoto et al. / Magnetic Resonance Imaging 33 (2015) 644–648

obtained in such cases are not sufficiently reliable. In this context, by using the T2-corrected internal reference method that we reported, the T2 relaxation curve can be derived from three TE points, and the signal for TE = 0 ms can be obtained by extrapolation. This makes it possible, under all TE settings, to obtain quantitative values close to the actual T2-corrected values [11,12]. The objective of this study is to elucidate how using TE that is longer than recommended for LCModel affects the quantitative values; this is accomplished by comparing the quantitative values for metabolite concentration calculated by using LCModel and those calculated by using the T2-corrected internal reference method. 2. Material and methods Proton MRS was conducted with 5 healthy volunteers and 11 patients with high-grade glioma (4 with anaplastic oligoastrocytoma, 1 with ependymoma and 6 with glioblastoma). The apparatus used was a 3 T MR system (Philips Achieba) with a sensitivity-encoding head coil. Point-resolved spectroscopy (PRESS) sequence was used for region selection. The following imagining parameters were used: TR of 6000 ms; TE of 72, 144 or 288 ms; bandwidth of 2000 Hz; sampling of 1024; and number of excitations of 128. The chemical shift selective model was used for water suppression, and shimming was conducted automatically. Using LCModel and the T2-corrected internal reference method [10,11], we quantified the concentrations of the metabolites N-acetylaspartate (NAA), creatine and phosphocreatine (Cr), and choline-containing compounds (Cho) and compared the results. The Stell–Dwass method was used to test the results for significant differences. For the spectral analysis using the T2-corrected internal reference method, we used a Java-based graphical user interface for magnetic resonance (http://sermn02.uab.es/mrui/) [13,14]. The adopted T2 correction internal standard method for signal intensity is described below. First, the TE is set at 3 points (TE 72, 144, and 288 ms)

645

in order to acquire spectra with and without water suppression. Next, the T2 relaxation time of water and metabolites is determined from Eq. (1), followed by correction of their respective signal intensity.   −2τ MS ¼ M 0  exp T2

Eq: ð1Þ

where MS denotes the signal intensity of the measurement at a given TE, M0 denotes the signal intensity at a TE of 0, and τ denotes TE/2. Previous research [11] has revealed that T1 relaxation has a small effect on quantitative values (1/15 of the effect of T2 relaxation); in this study, we therefore examined only T2 relaxation. This study was approved by the ethics committee at our hospital. The purpose and meaning of the study were explained to the participants, and their written informed consent was obtained. 3. Results Fig. 1 shows the relation between TE and the ratios of quantitative values obtained using LCModel and the T2-corrected internal reference method (LCModel/T2correct) for healthy participants. The figure shows the influence of changing TE on the ratios of quantitative values obtained using LCModel and the T2-corrected internal reference method for the five healthy participants. The vertical axes denote the ratio of the quantitative values (LCModel/T2correct), and the horizontal axes denote TE. The ratios increased with increasing TE (p b 0.05), and in the case of NAA at TE = 288 ms, the quantitative values obtained using LCModel were overestimated by a factor of 6.8. Furthermore, the ratios of the values from the two models differed between metabolites (p b 0.05). Fig. 2 shows the relation between TE and the ratios of quantitative values obtained using LCModel and the T2-corrected internal reference model (LCModel/T2correct) for patients with

Fig. 1. Influence of changes in TE on the quantitative values obtained using LCModel (healthy volunteers).

646

T. Yamamoto et al. / Magnetic Resonance Imaging 33 (2015) 644–648

Fig. 2. Influence of changing TE on the quantitative value of LCModel (patients with glioma).

high-grade glioma. Although the ratios of quantitative values obtained using the two models were found to increase with increasing TE, there was no significant difference since the ratios show large differences between individuals. Analysis was not possible in about 80% of the cases for NAA concentrations calculated using the T2-corrected internal reference method because spectral data could not be acquired due to a weak signal at TE = 288 ms and the reduced amount of NAA in brain tumors. Fig. 3 shows the spectrum of glioma acquired using MRUI (a) and LCModel (b). Both are of sufficient quality for the analysis of metabolite concentrations.

4. Discussion We calculated the ratios of quantitative values by using LCModel and the T2-corrected internal reference method (LCModel/T2correct) and examined how the quantitative results were influenced by TE that was longer than recommended for LCModel. In healthy volunteers, longer TE values yielded larger ratios of quantitative values for both LCModel and the T2-corrected internal reference method, with the ratio reaching 6.8 at TE = 288 ms in the case of NAA (p b 0.05). In other words, LCModel overestimated the values as compared with the T2-corrected internal reference method. Furthermore, the ratio of quantitative values from the two models differed substantially between metabolites (p b 0.05). In patients with high-grade glioma, although the ratios tended to increase with increasing TE, there was no significant difference because the ratios showed substantial differences between individuals. The quantitative values for LCModel were calculated using the water scaling method [5], which takes water in the tissue as a reference, similarly to the T2-corrected internal reference method. However, the two models use different techniques for relaxation time correction of signal intensity. In the T2-corrected internal reference method, the signal at TE = 0 is extrapolated from signals acquired at several values of TE. In contrast, LCModel calculates the signal value at TE = 0 by using a correction coefficient, which is calculated assuming that data are acquired under the recommended settings (TR of 6000 ms or greater, TE of 30 ms or less).

When TE was longer than recommended, the ratios of quantitative values obtained using LCModel and the T2-corrected internal reference method were vastly different for different metabolites (Fig. 1). This stems from the water scaling method (Eq. (2)). C metabolite ¼ C water 

    2 S  metabolite n Swater

Eq: ð2Þ

Cmetabolite: concentration of metabolite (mmol/kg wet weight); Cwater: concentration of water in tissue (fixed: 35 mol/kg wet weight) [15,16]; Smetabolite: signal intensity for metabolite; Swater: signal intensity for water in tissue; n: number of protons contributing to the signal. As seen in Eq. (2), quantitative values obtained using LCModel are related to the concentration of water in the tissue (Cwater), the number of protons related to the resonance (n) and the ratio of the signal intensities for metabolite and water (Smetabolite/Swater). Among these, Cwater and n are not influenced by TE, which is the reason why Smetabolite/Swater determines the quantitative value when TE is changed. When TE is changed, Smetabolite/Swater is influenced by T2 relaxation. If TE is short, signal attenuation due to T2 relaxation is small. In other words, even if the T2 relaxation time differs between metabolites, the corresponding changes in ratios of the signals for metabolites and water (i.e., NAA/water, Cr/water and Cho/water), which are caused by signal attenuation, are small. In contrast, if TE is long, signal attenuation due to T2 relaxation becomes large. Therefore, if the T2 relaxation time differs between metabolites, there are large variations in the ratios of the signal intensities for metabolite and water due to signal attenuation. In this regard, Fig. 4 shows a simulation of the relation between TE and the signal intensity ratios (NAA/water and Cr/water). The respective relaxation times for NAA and Cr are 370 and 200 ms [10]. Assuming that the signal intensities are the same at TE = 0 ms, the ratios of Cr/NAA (i.e., Cr/water: NAA/water) at TE = 72 ms, 144 ms and 288 ms are 1:1.2, 1:1.4 and 1:1.9, respectively. In other words, when TE is short, the influence of the differences in T2 relaxation time between metabolites on the quantitative value is small, however, it increases proportionally to TE. Based on the above, the difference between the ratios of quantitative values for metabolites as obtained using LCModel and the T2-corrected

T. Yamamoto et al. / Magnetic Resonance Imaging 33 (2015) 644–648

647

Fig. 3. MR spectrum of glioma acquired using MRUI (a) and LCModel (b). The spectrum were obtained by different TEs(72, 144, 288 ms).

internal reference method is considered to originate from the differences in T2 relaxation time between metabolites. In the case of healthy volunteers and long TE, the quantitative values for metabolite concentration obtained using LCModel were larger than those obtained using the T2-corrected internal reference method (Fig. 1). This is attributed to the influence of signal attenuation caused by the prolonged TE. LCModel uses a coefficient to compensate for the influence of signal attenuation caused by T2 relaxation. However, the LCModel correction coefficient is calculated for the recommended setting (TE = 30 ms or less). Furthermore, the initial settings for LCModel include T2 relaxation

Fig. 4. Influence of changing TE on signal intensity ratio (NAA/water and Cr/water).

correction for only water in the tissue. Fig. 5 shows a simulation of T2 relaxation correction with the correction coefficient in LCModel performed by using the signal for water in the tissue obtained at TE = 288 ms, which is longer than the recommended TE value. Due to the influence of T2 relaxation, signal attenuation is more pronounced at TE = 288 ms than at TE = 30 ms. However, even at TE = 288 ms, LCModel performs signal attenuation correction using the correction coefficient obtained under the recommended settings, resulting in insufficient correction. As mentioned above, quantitative values calculated using LCModel are influenced by the ratio of the signal intensities of metabolite and water (Smetabolite/Swater) (Eq. (2)). If the correction of the signal for water in the tissue is insufficient, the ratio of the signal intensities of metabolite and water becomes large

Fig. 5. Influence of the LCModel correction coefficient on T2 relaxation correction (water in tissue).

648

T. Yamamoto et al. / Magnetic Resonance Imaging 33 (2015) 644–648

accordingly. For this reason, LCModel overestimates quantitative values in comparison with the T2-corrected internal reference method. Based on the above, when TE is longer than recommended, it is necessary to adjust the correction coefficient since the signal correction is insufficient. Although the ratios of quantitative values obtained using LCModel and the T2-corrected internal reference method tended to be large at long TE, there was no significant difference because the ratios differed substantially between patients (Fig. 2), owing to the large differences in T2 relaxation times of water and metabolites between individual patients [10,17–19]. In the T2-corrected internal reference method, the signal value at TE = 0 ms is extrapolated by deriving the T2 relaxation curve from three TE points. For this reason, even if the T2 relaxation times for water and metabolites are different between individual patients, it is possible to correct the signal value and bring it closer to the actual value [11]. In contrast, in the case of LCModel, the correction coefficient must be adjusted for each patient to perform T2 relaxation correction. Although the correction coefficient can be changed by the user, T2 relaxation times, which are different for each patient, cannot be accurately obtained from data including only a single TE point. For this reason, accurate T2 relaxation correction is considered infeasible when using only the correction coefficient in clinical practice. For patients whose T2 relaxation time is expected to deviate substantially from the range of accuracy, caution must be exercised when interpreting quantitative values obtained using TE that is longer than recommended. A previous report examined the effect of relaxation time in spectrum analysis using LCModel in subjects with mesial temporal sclerosis [20]. In this report, an analysis of the data obtained with short TE and long TE was carried out using LCModel; however, a difference in metabolite concentration was not seen, in contrast to our results. The reason for this is three points. First, the studies differed in their subjects (diseases). Jírů F et al. targeted patients with nontumoral condition, whereas our study targeted patients with brain tumors. The changes of T2 relaxation time in patients with brain tumors are relatively larger compared to those in patients with non-tumoral condition. Second was the difference in sequence. Jírů F et al. acquired their data for a short TE using stimulated echo acquisition mode (STEAM) sequence and for a long TE using PRESS sequence, whereas in our study, the data were obtained using PRESS sequence for both the short and long TE. Third was the difference in magnetic field strength. Jírů F et al. used 1.5 T MR equipment, while we used 3 T MR equipment. Theoretically, differences of T2 relaxation time can be more sensitively detected with higher magnetic field strength; thus, this may have led to our clearly distinct results. 5. Conclusions We investigated how TE in spectral data acquisition influences quantitative values obtained using LCModel. In clinical practice, TE is sometimes set longer than the value recommended for LCModel. If

TE is long, LCModel overestimates the quantitative value since it cannot compensate for signal attenuation, and this effect is different for each metabolite and condition. Therefore, if TE is longer than recommended, it is necessary to account for the possibly reduced reliability of quantitative values calculated using LCModel.

References [1] Demaerel P, Johannik K, Van Hecke P, Van Ongeval C, Verellen S, Marchal G, et al. Localized 1H NMR spectroscopy in fifty cases of newly diagnosed intracranial tumors. J Comput Assist Tomogr 1991;15:67–76. [2] Shimizu H, Kumabe T, Tominaga T, Kayama T, Hara K, Ono Y, et al. Noninvasive evaluation of malignancy of brain tumors with proton MR spectroscopy. AJNR 1996;17:737–47. [3] Michaelis T, Merboldt KD, Bruhn H, Hänicke W, Frahm J. Absolute concentrations of metabolites in the adult human brain in vivo: quantification of localized proton MR spectra. Radiology 1993;187:219–27. [4] Provencher SW. Automatic quantitation of localized in vivo 1H spectra with LCModel. NMR Biomed 2001;14:260–4. [5] Provencher SW. Estimation of metabolites concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993;30:672–9. [6] Isobe T, Matsumura A, Anno I, Kawamura H, Muraishi H, Umeda T, et al. Effect of J coupling and T2 relaxation in assessing of methyl lactate signal using PRESS Sequence MR spectroscopy. Jpn J Med Phys 2005;25:68–74. [7] Isobe T, Matsumura A, Anno I, Yoshizawa T, Itai Y, Nose T. Proton MRS in a case of intracerebellar epidermoid. Brain Merve 2002;54:163–6. [8] Isobe T, Matsumura A, Anno I, Kawamura H, Shibata Y, Muraishi H, et al. Lactate quantification by proton magnetic resonance spectroscopy using a clinical MRI machine: a basic study. Australas Radiol 2007;51:330–3. [9] Poptani H, Gupta RK, Roy R, Pandey R, Jain VK, Chhabra DK. Characterization of intracranial mass lesions with in vivo proton MR spectroscopy. AJNR Am J Neuroradiol 1995;16:1593–603. [10] Isobe T, Matsumura A, Anno I, Yoshizawa T, Nagatomo Y, Itai Y, et al. Quantification of cerebral metabolites in glioma patients with proton MR spectroscopy using T2 relaxation time correction. Magn Reson Imaging 2002;20:343–9. [11] Isobe T, Akutsu H, Shiigai M, Masumoto T, Tadano K, Hirano Y, et al. Change of relaxation time and quantification of brain metabolites in proton MR spectroscopy. Jpn J Magn Reson Med 2010;30:122–33. [12] Frahm J, Bruhn H, Gyngell ML, Merboldt KD, Hänicke W, Sauter R. Localized proton NMR spectroscopy in different regions of the human brain in vivo. Relaxation times and concentrations of cerebral metabolites. Magn Reson Med 1989;11:47–63. [13] Naressi A, Couturier C, Devos JM, Janssen M, Mangeat C, de Beer R, et al. Javabased graphical user interface for the MRUI quantitation package. MAGMA 2001; 12:141–52. [14] Naressi A, Couturier C, Castang I, de Beer R, Graveron-Demilly D. Java-based graphical user interface for MRUI, a software package for quantitation of in vivo/medical magnetic resonance spectroscopy signals. Comput Biol Med 2001;31:269–86. [15] Ernst T, Kreis R, Ross B. Absolute quantitation of water and metabolites in the human brain. I: Compartments and water. J Magn Reson B 1993;102:1–8. [16] Christiansen P, Toft PB, Gideon P, Danielsen ER, Ring P, Henriksen O. MR-visible water content in human brain: a proton MRS study. Magn Reson Imaging 1994; 12:1237–44. [17] Ott D, Hennig J, Ernst T. Human brain tumors: assessment with in vivo proton MR spectroscopy. Radiology 1993;186:745–52. [18] Zeng QS, Li CF, Zhang K, Liu H, Kang XS, Zhen JH. Multivoxel 3D proton MR spectroscopy in the distinction of recurrent glioma from radiation injury. J Neurooncol 2007;84:63–9. [19] Szigety SK, Allen PS, Huyser-Wierenga D, Urtasun RC. The effect of radiation on normal human CNS as detected by NMR spectroscopy. Radiat Oncol 1993;25: 695–701. [20] Jírů F, Dezortová M, Burian M, Hájek M. The role of relaxation time corrections for the evaluation of long and short echo time 1H MR spectra of the hippocampus by NUMARIS and LCModel techniques. MAGMA 2003;16:135–43.

Influence of echo time in quantitative proton MR spectroscopy using LCModel.

The objective of this study was to elucidate the influence on quantitative analysis using LCModel with the condition of echo time (TE) longer than the...
719KB Sizes 0 Downloads 6 Views