ORIGINAL RESEARCH
Scan–Rescan Reproducibility of Parallel Transmission Based Amide Proton Transfer Imaging of Brain Tumors Osamu Togao, MD, PhD,1* Akio Hiwatashi, MD, PhD,1 Jochen Keupp, PhD,2 Koji Yamashita, MD, PhD,1 Kazufumi Kikuchi, MD, PhD,1 Takashi Yoshiura, MD, PhD,1 Yuriko Suzuki,3 Marijn J. Kruiskamp, PhD,2 Koji Sagiyama, MD,1,4 Masaya Takahashi, PhD,4 and Hiroshi Honda, MD, PhD1 Purpose: To evaluate the reproducibility of amide proton transfer (APT) imaging of brain tumors using a parallel transmission-based technique. Materials and Methods: Thirteen patients with brain tumors (four low-grade gliomas, three glioblastoma multiforme, five meningiomas, and one malignant lymphoma) were included in the study. APT imaging was conducted at 3T using a 2-channel parallel transmission scheme with a saturation time of 2 seconds and B1 amplitude of 2 lT. A 2D fast spinecho sequence with driven-equilibrium refocusing was used for imaging. Z-spectra were obtained at 25 frequency offsets from 26 to 16 ppm (step 0.5 ppm). A point-by-point B0 correction was performed with a B0 map. A scan–rescan reproducibility test was performed in two sessions on separate days for each patient. The interval between the two sessions was 4.8 6 3.5 days. Regions-of-interest (ROIs) were placed to include the whole tumor for each case. A mean and 90-percentile value of APT signal for the whole tumor histogram was calculated for each session. The between-session and within-session reproducibility was evaluated using linear regression analysis, intraclass correlation coefficient (ICC), and a Bland-Altman plot. Results: The mean and 90-percentile values of the APT signal for whole tumor ROI showed excellent agreements between the two sessions, with R2 of 0.91 and 0.96 in the linear regression analysis and ICC of 0.95 and 0.97, respectively. Conclusion: Parallel transmission-based APT imaging of brain tumors showed good reproducibility. J. MAGN. RESON. IMAGING 2015;42:1346–1353.
A
mide proton transfer (APT) imaging is one subset of the endogenous chemical exchange saturation transfer (CEST) imaging techniques introduced by the group of Zhou and van Zijl et al.1,2 APT imaging leverages the proton exchange between the amide protons (–NH) in endogenous mobile proteins and bulk-water protons.1 Thus, it is thought that the APT signal quantified by the asymmetry of the magnetization transfer (MT) at 3.5 ppm relative to the bulk water reflects concentrations of amide containing proteins or the exchange rate of amide protons. APT imaging is considered the most clinically applicable technique among various CEST imaging methods that have emerged in the
field of molecular imaging. It is with this thought that, first, APT imaging is available on clinical 3T scanners without special hardware modifications. Second, APT contrast is endogenous and no administration of a contrast agent is required. Third, the offset frequency of amide protons (13.5 ppm) is relatively far from the bulk water compared with endogenous CEST techniques of hydroxyl protons (– OH) (0.5–1.5 ppm).3,4 APT imaging is expected to play an important role in characterization of brain tumors 5,6 as well as of extracranial tumors.7 Recently, Togao et al8 demonstrated that APT imaging at 3T was able to predict histopathological grades of diffuse gliomas in patients. It has also
View this article online at wileyonlinelibrary.com. DOI: 10.1002/jmri.24895 Received Jan 12, 2015, Accepted for publication Mar 9, 2015. *Address reprint requests to: O.T., Department of Clinical Radiology, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi Higashi-ku Fukuoka, Japan, 812-8582. E-mail: [email protected] From the 1Department of Clinical Radiology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan; 2Philips Research Europe, Hamburg, Germany; 3Philips Electronics Japan, Tokyo, Japan; and 4Advanced Imaging Research Center, UT Southwestern Medical Center, Dallas, Texas, USA
1346
C 2015 Wiley Periodicals, Inc. V
Togao et al.: Reproducibility of APT Imaging
FIGURE 1: Scan–rescan protocol design. APT imaging was performed in two sessions on separate days for each patient. The interval between the two sessions was 4.8 6 3.5 days (2–14 days). In the second session, an additional short scan without repositioning was performed to assess reproducibility in a single session.
been shown that APT imaging could be useful in the assessment of therapeutic response to chemotherapy of glioblastoma multiforme (GBM),9 and in differentiation of radiation necrosis and active/recurrent glioma.10 Although these pioneering works demonstrated the potential of APT imaging to have a substantial impact on patient management, it is extremely important that the reproducibility be established as a quantitative and reliable imaging method in clinical use. It is known that quantification of APT signal is sensitive to accuracy of saturation pulse (B1) and B0 inhomogeneity. Recently, a technique based on a parallel radiofrequency (RF) transmission has been developed,11 which allows for arbitrarily long (5 sec) and constant RF pulses via amplifier alternation in clinical scanners. Therefore, this technique combined with a special RF shimming scheme for B1 homogeneity may improve the technical concerns in APT imaging. The purpose of this study was to assess the reproducibility of parallel transmission based APT imaging of brain tumors by means of a scan–rescan test.
MATERIALS AND METHODS Patients This study was approved by the Institutional Review Board and informed consent was obtained. Thirteen patients with brain tumors (49.3 6 23.0 years old, eight males and five females) who underwent subsequent surgical resection were included in the prospective study. Histological types of brain tumors were as follows: four low-grade gliomas, three GBMs, five meningiomas, and one malignant lymphoma.
Protocol of Scan–Rescan Test Figure 1 shows a schema of the scan–rescan protocol. APT imaging was performed in two sessions on separate days for each patient to assess between-session reproducibility. The interval between the two sessions was 4.8 6 3.5 days (2–14 days). In the second session, an additional short scan without repositioning was performed to assess within-session reproducibility.
MRI MRI was performed on a 3T clinical scanner (Achieva 3.0TX, Philips Healthcare, Best, The Netherlands) equipped with a secondNovember 2015
order shim, using an 8-channel head coil for signal reception and 2channel parallel transmission via the body coil for RF transmission. The acquisition software was modified to alternate the operation of the two transmission channels during the RF saturation pulse. The alternate activation of the two transmission channels enables long quasicontinuous RF saturation beyond the 50% duty-cycle of a single RF amplifier.11 Parallel transmission-based APT imaging functionally allows saturation pulse lengths of up to 5 seconds with the full available power of one RF amplifier to evoke the maximum B1 field amplitude. Since all imaging pulses within a parallel transmission-based APT sequence are using both amplifiers together in a standard way, there are no restrictions regarding the choice of MR sequence types (spin-echo/gradient-echo), because the full RF power range is available. The two RF channels in the parallel transmission system were used with the same pulse shape of single lobe sinc-Gaussian pulse. The RF amplitudes were adapted per channel to achieve similar mean B1 field amplitudes from both channels. The saturation builds up over the duration of the pulse train, and thus, the amplitudes can be used for B1 shimming (phase independent addition of the two channel contributions). The acquisition software was modified to allow a special RF shimming for the saturation homogeneity of the alternated saturation pulse.11 Scan geometry planning was performed using the function (SmartExam) on the MR console that allows for fully automated and consistent planning based on landmark detection of the brain.12 This ensured consistent slice selections on the two separate sessions in each subject. Following the second-order B0 shimming, 2D APT imaging was performed on a single slice corresponding to a maximum cross-section area of a tumor, using a saturation pulse with a duration of 2 seconds (50 msec 3 40 elements) with sincGaussian-shaped pulses and a saturation power level corresponding to B1 5 2 lT. To obtain a Z-spectrum, imaging was repeated at 25 saturation frequency offsets from x 5 26 to 16 ppm with a step of 0.5 ppm as well as one far off-resonant frequency (x 5 2160 ppm) for signal normalization. Another shortened version of APT imaging (scan time 5 35 sec) was performed at seven frequency offsets (24.0, 23.5, 23.0, 13.0, 13.5, 14.0, 2160 ppm). The other imaging parameters were as follows: fast spin-echo readout with driven equilibrium refocusing; echo train length 128 (singleshot fast spin-echo); sensitivity encoding (SENSE) factor 2; repetition time (TR) 5 5000 msec; echo time (TE) 5 6 msec; matrix 5 128 3 128 (reconstructed to 256 3 256); slice thickness 5 5 mm, field of view 5 230 3 230 mm; scan time 5 2 minutes 20 seconds for one Z-spectrum with 25 saturation frequency offsets. A B0 map for off-resonance correction was acquired separately (2D gradient-echo sequence, TR 5 15 msec, TE 5 1 and 2 msec, dual echo, 16 averages, 33 sec) in the same imaging geometry for the APT imaging, and it was used for a point-by-point B0 inhomogeneity correction. For reference, several standard MR images, including T1-weighted, T2-weighted, fluid-attenuated inversion recovery (FLAIR), and contrast-enhanced T1-weighted images were acquired. The following parameters were used: T2weighted (2D fast spin-echo, TR/TE 5 3000/80 msec, echo train length 5 8, 18 slices, thickness 5 5 mm, SENSE factor 5 1.6); preand postcontrast T1-weighted (2D spin-echo sequence, TR/ TE 5 500/10 msec, 18 slices, thickness 5 5 mm); FLAIR (TR/TE/ inversion time 5 10,000/100/2800 msec, echo train length 5 27, 1347
Journal of Magnetic Resonance Imaging
FIGURE 2: A case of GBM (grade IV) in the left thalamus. (A) Good between-scan and within-session reproducibility was observed. Note that the APT-weighted images in the two sessions on separate days are almost identical, although they were obtained with different positioning and B0 shimming. (B) The quantile-quantile plots of the APT signal for both within-session and betweensession analyses show linear relationship between the two scans, although some deviations are seen in higher APT signals. Both analyses show a similar distribution of APT signals within the tumor.
18 slices, thickness 5 5 mm, SENSE factor 5 1.6). The APT images were acquired before the administration of the gadolinium contrast agent in all patients.
APT Image Data Analysis APT imaging analysis was performed with the software program ImageJ (v. 1.43u; National Institutes of Health, Bethesda, MD). Histogram analysis was performed with the software program Medical Image Processing, Analysis and Visualization (v. 4.1.1; NIH, Bethesda, MD). A dedicated ImageJ plug-in was build to analyze the Z-spectra and asymmetry of magnetization transfer ratio (MTRasym) equipped with a correction function for B0 inhomogeneity, using interpolation among the Z-spectral image data. The local B0 field shift in Hz was obtained from the B0 map, which was created from dual-echo gradient-echo images (TE 5 1 and 2 msec) according to the following equation: DB0(x) 5 (Phase[TE2](x)-Phase[TE1](x)) / (TE2-TE1)*2*Pi, where phase [TEi](x) indicates image phases of the images with echo times TE1 or TE2 at position x in radians, and TE1 and TE2 are given in seconds. The DB0(x) is the resulting B0 map measured in Hz. Each voxel was corrected in image intensity for the nominal saturation frequency offset by Lagrange interpolation 1348
among the neighboring Z-spectral images. This procedure corresponds to a frequency shift along the saturation frequency offset axis according to the measured B0 shift. The Z-spectrum was calculated as Ssat/S0, where Ssat and S0 are the signal intensity obtained with and without selective saturation, respectively.1 To reduce these undesired contributions from conventional MT effect and direct saturation of bulk water, an asymmetry analysis of the Z-spectrum with respect to the water frequency was performed as MTRasym.
MTRasym 5
Ssat ð2offsetÞ2Ssat ð1offsetÞ S0
APT signal was defined as the asymmetry of the Z-spectrum at 3.5 ppm calculated as MTRasym (3.5 ppm).
APT signal5MTRasym ð3:5 ppmÞ5
Ssat ð3:5 ppmÞ2Ssat ð13:5 ppmÞ : S0
A region-of-interest (ROI) was manually drawn to include a whole tumor area on a raw image at 13.5 ppm by a trained Volume 42, No. 5
Togao et al.: Reproducibility of APT Imaging
FIGURE 3: A case of diffuse astrocytoma (grade II) in the left temporal lobe. The tumor shows lower APT signal compared with the patient with GBM (Fig. 2). The small signal variability in the tumor is reproducibly observed in the APT images.
neuroradiologist (O.T., 14 years of experience in neuroradiology) with reference to conventional MR images. The ROI was copied onto the APT-weighted image to measure APT signals in all voxels included in the whole tumor ROI. A mean and 90-percentile value of APT signals in the whole tumor ROI was obtained based on the histogram analysis for each scan.
Histologic Evaluation The pathological diagnosis was determined with specimens obtained at surgical resection according to the World Health Organization (WHO) criteria13 by established neuropathologists.
Statistical Analysis All values were expressed as mean 6 standard deviation (SD). The correlation of APT signals between two scans was assessed by a simple linear regression analysis. The agreement of the APT signals between two scans was assessed by the intraclass correlation coefficient (ICC)14 and Bland-Altman plot analysis.15 ICCs were considered excellent if greater than 0.74.14 The quantile-quantile plot analysis was performed with R (The R Foundation for Statistical Computing, Vienna, Austria, http://www.r-project.org). Other statistical analyses were performed with a commercially available software package (SPSS, IBM 19, Armonk, NY, or Prism 5.0,
FIGURE 4: A case of fibrous meningioma (grade I) in the left parietal region. The tumor consistently demonstrates low overall APT signals in all three scans. A small high signal intensity spot in the tumor is observed in all scans.
November 2015
1349
Journal of Magnetic Resonance Imaging
FIGURE 5: A case of primary central nerve system lymphoma in the left temporal and occipital lobes. The tumor consistently shows high overall APT signals in all three scans.
GraphPad Software, San Diego, CA). P-values < 0.05 were considered significant.
RESULTS Figure 2 shows a representative case of GBM, which showed a good between-session and within-session reproducibility. Note that the APT-weighted images (Fig. 2A) in the two sessions on separate days are almost identical although they were obtained at different positioning and B0 shimming. The quantile-quantile plots (Fig. 2B) of the APT signal for both within-session and between-session analyses show linear relationship between the two scans, although some deviations are seen in higher APT signals. Both analyses show similar distribution of APT signals within the tumor. Figure
3 is a representative case of low-grade glioma (diffuse astrocytoma, WHO grade II) in the left temporal lobe, which shows lower APT signal compared with the patient with GBM (Fig. 2). The small signal variability in the tumor is reproducibly observed in the APT-weighted images. Figures 4 and 5 are representative cases of meningioma and malignant lymphoma, respectively. Figure 6 shows the mean (Fig. 6A) and 90-percentile (Fig. 6B) value of the APT signal of a whole tumor ROI in the two separate sessions (between-session reproducibility). Excellent correlations between the two sessions were observed in the mean APT signal (R2 5 0.91, P < 0.0001) and in the 90-percentile APT signal (R2 5 0.92, P < 0.0001). Excellent agreements between the two sessions
FIGURE 6: Mean (A) and 90-percentile (B) value of the APT signal of a whole tumor ROI in the two sessions. Excellent correlations between the two sessions are observed in the mean APT signal (R2 5 0.91, P < 0.0001) and in the 90-percentile APT signal (R2 5 0.92, P < 0.0001).
1350
Volume 42, No. 5
Togao et al.: Reproducibility of APT Imaging
FIGURE 7: Bland-Altman plot analysis of the mean (A) and 90-percentile (B) of APT signal of tumor in the two sessions. The mean and 90-percentile APT signals show good accordance, with only one and two values beyond the 95% limits of agreement (broken lines), respectively. No additive or proportional biases were observed in either analyses.
were observed in the mean APT signal (ICC 5 0.95) as well as in the 90-percentile APT signal (ICC 5 0.95). Figure 7 shows the Bland-Altman plot analysis of the mean (Fig. 7A) and 90-percentile (Fig. 7B) of APT signal of tumor in the two sessions. The mean and 90-percentile APT signals showed excellent accordance, with one and two values beyond the 95% limits of agreement, respectively. No additive or proportional biases were observed in either analysis. Figure 8 shows the mean (Fig. 8A) and 90-percentile (Fig. 8B) value of the APT signal of a whole tumor ROI in the two scans in the single session without repositioning (within-session reproducibility). Excellent correlations between the two scans in a session were observed in the mean APT signal (R2 5 0.96, P < 0.0001) and in the 90percentile APT signal (R2 5 0.93, P < 0.0001). Excellent agreements between the two scans in the session were observed in the mean APT signal (ICC 5 0.98) and the 90percentile APT signal (ICC 5 0.97). Figure 9 shows the Bland-Altman plot analysis of the mean (Fig. 9A) and 90-
percentile (Fig. 9B) of APT signal of tumor in the two scans in the single session. The mean APT signals showed excellent accordance, with one value beyond the 95% limits of agreement. The 90-percentile APT signals showed excellent accordance, all values within the 95% limits of agreement. No additive or proportional biases were observed in either analyses.
DISCUSSION Endogenous CEST imaging suffers from small saturation transfer effects since the resonance frequency of the solute protons is sufficiently close to the bulk water frequency and the CEST effects have to compete with direct saturation of the water protons.16,17 The situation is even worse in vivo due to the large background MT effects. For example, in the study for grading diffuse gliomas, it was reported that APT SI was 2.1 6 0.4% in the grade II gliomas, 3.2 6 0.9% in the grade III gliomas, and 4.1 6 1.0% in the grade IV gliomas.8 In order to sensitively detect such small signal differences, APT imaging must be consistent and reproducible.
FIGURE 8: Mean (A) and 90-percentile (B) value of the APT signal of a whole tumor ROI in the two scans in the single session without repositioning. Excellent correlations between the two scans in a session are observed in the mean APT signal (R2 5 0.96, P < 0.0001) and in the 90-percentile APT signal (R2 5 0.93, P < 0.0001).
November 2015
1351
Journal of Magnetic Resonance Imaging
FIGURE 9: Bland-Altman plot analysis of the mean (A) and 90-percentile (B) of APT signal of tumor in the two scans in the single session. The mean and 90-percentile APT signals show excellent accordance, with only one and no value beyond the 95% limits of agreement (broken line), respectively. No additive or proportional biases were observed in either analysis.
Spatial homogeneity of the saturation pulse is essential to ensure consistent APT imaging. Parallel transmission based APT imaging incorporated special RF shimming, which enables homogenized (adaptively shimmed) RF saturation pulses by adjusting the individual power of the RF amplifiers during alternation, according to the results of the RF calibration scan (B1 calibration). A previous study conducted in a pasteurized, homogeneous chicken egg-white phantom showed that gradient-echo-based APT image without RF shimming was spatially heterogeneous with 3% locational variation in APT signal in the egg-white phantom.11 On the other hand, the gradient-echo based APT image with RF shimming showed much more homogenous APT signal in the phantom with 1% locational variation. Furthermore, FSE with driven equilibrium sequence with RF shimming, which was used in this study, showed even more homogenous saturation effect with only 0.5% locational variation in APT signal, compared with the gradientecho sequence.11 Another critical point for consistent APT imaging is B0 inhomogeneity correction. Endogenous CEST imaging is extremely sensitive to magnetic field inhomogeneity.16,17 Due to the steep slope of the direct saturation curve of the water, even a small B0 field difference and a resulting shift in the Z-spectrum may cause a large change in MTRasym, where the local magnetic susceptibility effects are commonplace. In vivo, this effect may result in the occurrence of signal variation in APT images. To reduce this problem, image acquisition techniques have been performed that included many saturation frequency offsets, followed by polynomial or cubic-spline fitting and centering of the Z-spectrum in each voxel.2 This B0 correction method was applicable in APT imaging because the frequency offset of amide protons (13.5 ppm) is relatively large, eg, compared to hydroxyl protons (1 ppm). However, this correction approach is sometimes problematic when the MT effect of the tissue is large and the Z-spectrum is broad, in which the center of water frequency is difficult to determine accurately. In addi1352
tion, the conventional MT effect in the normal brain has been shown to be asymmetric.1,18 As an alternative, a B0 map obtained from the phase images of a dual-echo gradient-echo sequence was used in this study. B0 shift correction using a B0 map is a simple and robust method since this centering method is not affected by the MT effect and direct saturation of water. In this study, the mean and 90percentile of APT signal of tumor showed excellent agreement between the two sessions (between-session, B0 inhomogeneity corrected with different B0 maps), which was comparable to the agreements between the two scans in a single session (within-session, B0 inhomogeneity corrected with the same B0 map). These results indicate that B0 mapbased B0 inhomogeneity correction is accurate and reproducible. Another advantage of the use of a B0 map is that it is not necessary to obtain a full Z-spectrum and thus scan time can be considerably shortened. One scan in the second session in this study was obtained using only seven saturation frequency offsets around 63.5 ppm, which yielded almost identical APT-weighed maps to the other scan obtained from the full Z-spectrum acquisition (25 saturation frequency offsets) in spite of the shorter acquisition time (2 min 20 sec vs. 35 sec). The limitations of the study are as follows. The number of patients for each type of tumor and the total number of patients involved are small. There is no real reference standard for APT signal. Only a single observer drew an ROI; however, observer bias should have been small since the ROI was placed to include the whole tumor area. The slope in the between-session comparison deviated from 1 by 12% in mean APT signal and 16% in 90-percentile APT signal, whereas that in the within-session comparison was nearly 1. These deviations in between-session analysis can be attributed to difficulty in positioning the single slice to the exact same location and some variability in B0 shimming. The use of a single slice acquisition to reduce total scan time, since the APT imaging was included in the clinical preoperative scans for the patients. The multislice 2D and Volume 42, No. 5
Togao et al.: Reproducibility of APT Imaging 6.
Wen Z, Hu S, Huang F, et al. MR imaging of high-grade brain tumors using endogenous protein and peptide-based contrast. Neuroimage 2010;51:616–622.
7.
Togao O, Kessinger CW, Huang G, et al. Characterization of lung cancer by amide proton transfer (APT) imaging: an in-vivo study in an orthotopic mouse model. PLoS One 2013;8:e77019.
8.
Togao O, Yoshiura T, Keupp J, et al. Amide proton transfer imaging of adult diffuse gliomas: correlation with histopathological grades. Neurol Oncol 2014;16:441–448.
9.
Sagiyama K, Mashimo T, Togao O, et al. In vivo chemical exchange saturation transfer imaging allows early detection of a therapeutic response in glioblastoma. Proc Natl Acad Sci U S A 2014;111:4542– 4547.
10.
Zhou J, Tryggestad E, Wen Z, et al. Differentiation between glioma and radiation necrosis using molecular magnetic resonance imaging of endogenous proteins and peptides. Nat Med 2011;17:130–134.
11.
Keupp J, Baltes C, Harvey P, Van den Brink J. Parallel RF transmission based MRI technique for highly sensitive detection of amide proton transfer in the human brain at 3T. In: Proc 19th Annual Meeting ISMRM, Montreal; 2011.
12.
Petersen ET, Mouridsen K, Golay X. The QUASAR reproducibility study. Part II. Results from a multi-center arterial spin labeling testretest study. Neuroimage 2010;49:104–113.
13.
Louis DN, Ohgaki H, Wiestler OD, et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 2007; 114:97–109.
14.
Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86:420–428.
15.
Zhou J, Payen JF, Wilson DA, Traystman RJ, van Zijl PC. Using the amide proton signals of intracellular proteins and peptides to detect pH effects in MRI. Nat Med 2003;9:1085–1090.
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307– 310.
16.
2.
Zhou J, Lal B, Wilson DA, Laterra J, van Zijl PC. Amide proton transfer (APT) contrast for imaging of brain tumors. Magn Reson Med 2003; 50:1120–1126.
Zhou J, Blakeley JO, Hua J, et al. Practical data acquisition method for human brain tumor amide proton transfer (APT) imaging. Magn Reson Med 2008;60:842–849.
17.
3.
van Zijl PC, Jones CK, Ren J, Malloy CR, Sherry AD. MRI detection of glycogen in vivo by using chemical exchange saturation transfer imaging (glycoCEST). Proc Natl Acad Sci U S A 2007;104:4359–4364.
Kim M, Gillen J, Landman BA, Zhou J, van Zijl PC. Water saturation shift referencing (WASSR) for chemical exchange saturation transfer (CEST) experiments. Magn Reson Med 2009;61:1441–1450.
18.
4.
Ling W, Regatte RR, Navon G, Jerschow A. Assessment of glycosaminoglycan concentration in vivo by chemical exchange-dependent saturation transfer (gagCEST). Proc Natl Acad Sci U S A 2008;105:2266–2270.
Zhou J, Hong X, Zhao X, Gao JH, Yuan J. APT-weighted and NOEweighted image contrasts in glioma with different RF saturation powers based on magnetization transfer ratio asymmetry analyses. Magn Reson Med 2013;70:320–327.
5.
Salhotra A, Lal B, Laterra J, Sun PZ, van Zijl PC, Zhou J. Amide proton transfer imaging of 9L gliosarcoma and human glioblastoma xenografts. NMR Biomed 2008;21:489–497.
19.
Keupp J, Eggers H. CEST-DIXON MRI for sensitive and accurate measurement of amide proton transfer in humans at 3T. In: Proc 18th Annual Meeting ISMRM, Stockholm; 2010.
3D volume coverage acquisition with B0 inhomogeneity correction function is now under development. Possible subtle position shifts between the APT imaging and B0 map acquisition could have affected the quantitation of the APT signals. Efficient self-B0 correcting CEST methods using multiecho gradient-echo/fast spine-echo (CEST-DIXON) are expected to overcome this issue.19 In conclusion, we demonstrated a good reproducibility of the parallel transmission-based APT imaging of brain tumors confirmed by the scan–rescan test. Spatially homogenous RF shimming and accurate B0 inhomogeneity correction could have greatly contributed to the reproducibility.
Acknowledgment Contract grant sponsor: Japanese Society of Neuroradiology; Contract grant sponsor: Japanese Radiological Society; Contract grant sponsor: Philips Electronics Japan; Contract grant sponsor: Bayer Healthcare Japan; Contract grant sponsor: Fukuoka Foundation for Sound Health Cancer Research Fund; Contract grant numbers: Grant-in-Aid for Scientific Research numbers 26461827 and 22591340.
REFERENCES 1.
November 2015
1353
Scan–Rescan Reproducibility of Parallel Transmission Based Amide Proton Transfer Imaging of Brain Tumors Osamu Togao, MD, PhD,1* Akio Hiwatashi, MD, PhD,1 Jochen Keupp, PhD,2 Koji Yamashita, MD, PhD,1 Kazufumi Kikuchi, MD, PhD,1 Takashi Yoshiura, MD, PhD,1 Yuriko Suzuki,3 Marijn J. Kruiskamp, PhD,2 Koji Sagiyama, MD,1,4 Masaya Takahashi, PhD,4 and Hiroshi Honda, MD, PhD1 Purpose: To evaluate the reproducibility of amide proton transfer (APT) imaging of brain tumors using a parallel transmission-based technique. Materials and Methods: Thirteen patients with brain tumors (four low-grade gliomas, three glioblastoma multiforme, five meningiomas, and one malignant lymphoma) were included in the study. APT imaging was conducted at 3T using a 2-channel parallel transmission scheme with a saturation time of 2 seconds and B1 amplitude of 2 lT. A 2D fast spinecho sequence with driven-equilibrium refocusing was used for imaging. Z-spectra were obtained at 25 frequency offsets from 26 to 16 ppm (step 0.5 ppm). A point-by-point B0 correction was performed with a B0 map. A scan–rescan reproducibility test was performed in two sessions on separate days for each patient. The interval between the two sessions was 4.8 6 3.5 days. Regions-of-interest (ROIs) were placed to include the whole tumor for each case. A mean and 90-percentile value of APT signal for the whole tumor histogram was calculated for each session. The between-session and within-session reproducibility was evaluated using linear regression analysis, intraclass correlation coefficient (ICC), and a Bland-Altman plot. Results: The mean and 90-percentile values of the APT signal for whole tumor ROI showed excellent agreements between the two sessions, with R2 of 0.91 and 0.96 in the linear regression analysis and ICC of 0.95 and 0.97, respectively. Conclusion: Parallel transmission-based APT imaging of brain tumors showed good reproducibility. J. MAGN. RESON. IMAGING 2015;42:1346–1353.
A
mide proton transfer (APT) imaging is one subset of the endogenous chemical exchange saturation transfer (CEST) imaging techniques introduced by the group of Zhou and van Zijl et al.1,2 APT imaging leverages the proton exchange between the amide protons (–NH) in endogenous mobile proteins and bulk-water protons.1 Thus, it is thought that the APT signal quantified by the asymmetry of the magnetization transfer (MT) at 3.5 ppm relative to the bulk water reflects concentrations of amide containing proteins or the exchange rate of amide protons. APT imaging is considered the most clinically applicable technique among various CEST imaging methods that have emerged in the
field of molecular imaging. It is with this thought that, first, APT imaging is available on clinical 3T scanners without special hardware modifications. Second, APT contrast is endogenous and no administration of a contrast agent is required. Third, the offset frequency of amide protons (13.5 ppm) is relatively far from the bulk water compared with endogenous CEST techniques of hydroxyl protons (– OH) (0.5–1.5 ppm).3,4 APT imaging is expected to play an important role in characterization of brain tumors 5,6 as well as of extracranial tumors.7 Recently, Togao et al8 demonstrated that APT imaging at 3T was able to predict histopathological grades of diffuse gliomas in patients. It has also
View this article online at wileyonlinelibrary.com. DOI: 10.1002/jmri.24895 Received Jan 12, 2015, Accepted for publication Mar 9, 2015. *Address reprint requests to: O.T., Department of Clinical Radiology, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi Higashi-ku Fukuoka, Japan, 812-8582. E-mail: [email protected] From the 1Department of Clinical Radiology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan; 2Philips Research Europe, Hamburg, Germany; 3Philips Electronics Japan, Tokyo, Japan; and 4Advanced Imaging Research Center, UT Southwestern Medical Center, Dallas, Texas, USA
1346
C 2015 Wiley Periodicals, Inc. V
Togao et al.: Reproducibility of APT Imaging
FIGURE 1: Scan–rescan protocol design. APT imaging was performed in two sessions on separate days for each patient. The interval between the two sessions was 4.8 6 3.5 days (2–14 days). In the second session, an additional short scan without repositioning was performed to assess reproducibility in a single session.
been shown that APT imaging could be useful in the assessment of therapeutic response to chemotherapy of glioblastoma multiforme (GBM),9 and in differentiation of radiation necrosis and active/recurrent glioma.10 Although these pioneering works demonstrated the potential of APT imaging to have a substantial impact on patient management, it is extremely important that the reproducibility be established as a quantitative and reliable imaging method in clinical use. It is known that quantification of APT signal is sensitive to accuracy of saturation pulse (B1) and B0 inhomogeneity. Recently, a technique based on a parallel radiofrequency (RF) transmission has been developed,11 which allows for arbitrarily long (5 sec) and constant RF pulses via amplifier alternation in clinical scanners. Therefore, this technique combined with a special RF shimming scheme for B1 homogeneity may improve the technical concerns in APT imaging. The purpose of this study was to assess the reproducibility of parallel transmission based APT imaging of brain tumors by means of a scan–rescan test.
MATERIALS AND METHODS Patients This study was approved by the Institutional Review Board and informed consent was obtained. Thirteen patients with brain tumors (49.3 6 23.0 years old, eight males and five females) who underwent subsequent surgical resection were included in the prospective study. Histological types of brain tumors were as follows: four low-grade gliomas, three GBMs, five meningiomas, and one malignant lymphoma.
Protocol of Scan–Rescan Test Figure 1 shows a schema of the scan–rescan protocol. APT imaging was performed in two sessions on separate days for each patient to assess between-session reproducibility. The interval between the two sessions was 4.8 6 3.5 days (2–14 days). In the second session, an additional short scan without repositioning was performed to assess within-session reproducibility.
MRI MRI was performed on a 3T clinical scanner (Achieva 3.0TX, Philips Healthcare, Best, The Netherlands) equipped with a secondNovember 2015
order shim, using an 8-channel head coil for signal reception and 2channel parallel transmission via the body coil for RF transmission. The acquisition software was modified to alternate the operation of the two transmission channels during the RF saturation pulse. The alternate activation of the two transmission channels enables long quasicontinuous RF saturation beyond the 50% duty-cycle of a single RF amplifier.11 Parallel transmission-based APT imaging functionally allows saturation pulse lengths of up to 5 seconds with the full available power of one RF amplifier to evoke the maximum B1 field amplitude. Since all imaging pulses within a parallel transmission-based APT sequence are using both amplifiers together in a standard way, there are no restrictions regarding the choice of MR sequence types (spin-echo/gradient-echo), because the full RF power range is available. The two RF channels in the parallel transmission system were used with the same pulse shape of single lobe sinc-Gaussian pulse. The RF amplitudes were adapted per channel to achieve similar mean B1 field amplitudes from both channels. The saturation builds up over the duration of the pulse train, and thus, the amplitudes can be used for B1 shimming (phase independent addition of the two channel contributions). The acquisition software was modified to allow a special RF shimming for the saturation homogeneity of the alternated saturation pulse.11 Scan geometry planning was performed using the function (SmartExam) on the MR console that allows for fully automated and consistent planning based on landmark detection of the brain.12 This ensured consistent slice selections on the two separate sessions in each subject. Following the second-order B0 shimming, 2D APT imaging was performed on a single slice corresponding to a maximum cross-section area of a tumor, using a saturation pulse with a duration of 2 seconds (50 msec 3 40 elements) with sincGaussian-shaped pulses and a saturation power level corresponding to B1 5 2 lT. To obtain a Z-spectrum, imaging was repeated at 25 saturation frequency offsets from x 5 26 to 16 ppm with a step of 0.5 ppm as well as one far off-resonant frequency (x 5 2160 ppm) for signal normalization. Another shortened version of APT imaging (scan time 5 35 sec) was performed at seven frequency offsets (24.0, 23.5, 23.0, 13.0, 13.5, 14.0, 2160 ppm). The other imaging parameters were as follows: fast spin-echo readout with driven equilibrium refocusing; echo train length 128 (singleshot fast spin-echo); sensitivity encoding (SENSE) factor 2; repetition time (TR) 5 5000 msec; echo time (TE) 5 6 msec; matrix 5 128 3 128 (reconstructed to 256 3 256); slice thickness 5 5 mm, field of view 5 230 3 230 mm; scan time 5 2 minutes 20 seconds for one Z-spectrum with 25 saturation frequency offsets. A B0 map for off-resonance correction was acquired separately (2D gradient-echo sequence, TR 5 15 msec, TE 5 1 and 2 msec, dual echo, 16 averages, 33 sec) in the same imaging geometry for the APT imaging, and it was used for a point-by-point B0 inhomogeneity correction. For reference, several standard MR images, including T1-weighted, T2-weighted, fluid-attenuated inversion recovery (FLAIR), and contrast-enhanced T1-weighted images were acquired. The following parameters were used: T2weighted (2D fast spin-echo, TR/TE 5 3000/80 msec, echo train length 5 8, 18 slices, thickness 5 5 mm, SENSE factor 5 1.6); preand postcontrast T1-weighted (2D spin-echo sequence, TR/ TE 5 500/10 msec, 18 slices, thickness 5 5 mm); FLAIR (TR/TE/ inversion time 5 10,000/100/2800 msec, echo train length 5 27, 1347
Journal of Magnetic Resonance Imaging
FIGURE 2: A case of GBM (grade IV) in the left thalamus. (A) Good between-scan and within-session reproducibility was observed. Note that the APT-weighted images in the two sessions on separate days are almost identical, although they were obtained with different positioning and B0 shimming. (B) The quantile-quantile plots of the APT signal for both within-session and betweensession analyses show linear relationship between the two scans, although some deviations are seen in higher APT signals. Both analyses show a similar distribution of APT signals within the tumor.
18 slices, thickness 5 5 mm, SENSE factor 5 1.6). The APT images were acquired before the administration of the gadolinium contrast agent in all patients.
APT Image Data Analysis APT imaging analysis was performed with the software program ImageJ (v. 1.43u; National Institutes of Health, Bethesda, MD). Histogram analysis was performed with the software program Medical Image Processing, Analysis and Visualization (v. 4.1.1; NIH, Bethesda, MD). A dedicated ImageJ plug-in was build to analyze the Z-spectra and asymmetry of magnetization transfer ratio (MTRasym) equipped with a correction function for B0 inhomogeneity, using interpolation among the Z-spectral image data. The local B0 field shift in Hz was obtained from the B0 map, which was created from dual-echo gradient-echo images (TE 5 1 and 2 msec) according to the following equation: DB0(x) 5 (Phase[TE2](x)-Phase[TE1](x)) / (TE2-TE1)*2*Pi, where phase [TEi](x) indicates image phases of the images with echo times TE1 or TE2 at position x in radians, and TE1 and TE2 are given in seconds. The DB0(x) is the resulting B0 map measured in Hz. Each voxel was corrected in image intensity for the nominal saturation frequency offset by Lagrange interpolation 1348
among the neighboring Z-spectral images. This procedure corresponds to a frequency shift along the saturation frequency offset axis according to the measured B0 shift. The Z-spectrum was calculated as Ssat/S0, where Ssat and S0 are the signal intensity obtained with and without selective saturation, respectively.1 To reduce these undesired contributions from conventional MT effect and direct saturation of bulk water, an asymmetry analysis of the Z-spectrum with respect to the water frequency was performed as MTRasym.
MTRasym 5
Ssat ð2offsetÞ2Ssat ð1offsetÞ S0
APT signal was defined as the asymmetry of the Z-spectrum at 3.5 ppm calculated as MTRasym (3.5 ppm).
APT signal5MTRasym ð3:5 ppmÞ5
Ssat ð3:5 ppmÞ2Ssat ð13:5 ppmÞ : S0
A region-of-interest (ROI) was manually drawn to include a whole tumor area on a raw image at 13.5 ppm by a trained Volume 42, No. 5
Togao et al.: Reproducibility of APT Imaging
FIGURE 3: A case of diffuse astrocytoma (grade II) in the left temporal lobe. The tumor shows lower APT signal compared with the patient with GBM (Fig. 2). The small signal variability in the tumor is reproducibly observed in the APT images.
neuroradiologist (O.T., 14 years of experience in neuroradiology) with reference to conventional MR images. The ROI was copied onto the APT-weighted image to measure APT signals in all voxels included in the whole tumor ROI. A mean and 90-percentile value of APT signals in the whole tumor ROI was obtained based on the histogram analysis for each scan.
Histologic Evaluation The pathological diagnosis was determined with specimens obtained at surgical resection according to the World Health Organization (WHO) criteria13 by established neuropathologists.
Statistical Analysis All values were expressed as mean 6 standard deviation (SD). The correlation of APT signals between two scans was assessed by a simple linear regression analysis. The agreement of the APT signals between two scans was assessed by the intraclass correlation coefficient (ICC)14 and Bland-Altman plot analysis.15 ICCs were considered excellent if greater than 0.74.14 The quantile-quantile plot analysis was performed with R (The R Foundation for Statistical Computing, Vienna, Austria, http://www.r-project.org). Other statistical analyses were performed with a commercially available software package (SPSS, IBM 19, Armonk, NY, or Prism 5.0,
FIGURE 4: A case of fibrous meningioma (grade I) in the left parietal region. The tumor consistently demonstrates low overall APT signals in all three scans. A small high signal intensity spot in the tumor is observed in all scans.
November 2015
1349
Journal of Magnetic Resonance Imaging
FIGURE 5: A case of primary central nerve system lymphoma in the left temporal and occipital lobes. The tumor consistently shows high overall APT signals in all three scans.
GraphPad Software, San Diego, CA). P-values < 0.05 were considered significant.
RESULTS Figure 2 shows a representative case of GBM, which showed a good between-session and within-session reproducibility. Note that the APT-weighted images (Fig. 2A) in the two sessions on separate days are almost identical although they were obtained at different positioning and B0 shimming. The quantile-quantile plots (Fig. 2B) of the APT signal for both within-session and between-session analyses show linear relationship between the two scans, although some deviations are seen in higher APT signals. Both analyses show similar distribution of APT signals within the tumor. Figure
3 is a representative case of low-grade glioma (diffuse astrocytoma, WHO grade II) in the left temporal lobe, which shows lower APT signal compared with the patient with GBM (Fig. 2). The small signal variability in the tumor is reproducibly observed in the APT-weighted images. Figures 4 and 5 are representative cases of meningioma and malignant lymphoma, respectively. Figure 6 shows the mean (Fig. 6A) and 90-percentile (Fig. 6B) value of the APT signal of a whole tumor ROI in the two separate sessions (between-session reproducibility). Excellent correlations between the two sessions were observed in the mean APT signal (R2 5 0.91, P < 0.0001) and in the 90-percentile APT signal (R2 5 0.92, P < 0.0001). Excellent agreements between the two sessions
FIGURE 6: Mean (A) and 90-percentile (B) value of the APT signal of a whole tumor ROI in the two sessions. Excellent correlations between the two sessions are observed in the mean APT signal (R2 5 0.91, P < 0.0001) and in the 90-percentile APT signal (R2 5 0.92, P < 0.0001).
1350
Volume 42, No. 5
Togao et al.: Reproducibility of APT Imaging
FIGURE 7: Bland-Altman plot analysis of the mean (A) and 90-percentile (B) of APT signal of tumor in the two sessions. The mean and 90-percentile APT signals show good accordance, with only one and two values beyond the 95% limits of agreement (broken lines), respectively. No additive or proportional biases were observed in either analyses.
were observed in the mean APT signal (ICC 5 0.95) as well as in the 90-percentile APT signal (ICC 5 0.95). Figure 7 shows the Bland-Altman plot analysis of the mean (Fig. 7A) and 90-percentile (Fig. 7B) of APT signal of tumor in the two sessions. The mean and 90-percentile APT signals showed excellent accordance, with one and two values beyond the 95% limits of agreement, respectively. No additive or proportional biases were observed in either analysis. Figure 8 shows the mean (Fig. 8A) and 90-percentile (Fig. 8B) value of the APT signal of a whole tumor ROI in the two scans in the single session without repositioning (within-session reproducibility). Excellent correlations between the two scans in a session were observed in the mean APT signal (R2 5 0.96, P < 0.0001) and in the 90percentile APT signal (R2 5 0.93, P < 0.0001). Excellent agreements between the two scans in the session were observed in the mean APT signal (ICC 5 0.98) and the 90percentile APT signal (ICC 5 0.97). Figure 9 shows the Bland-Altman plot analysis of the mean (Fig. 9A) and 90-
percentile (Fig. 9B) of APT signal of tumor in the two scans in the single session. The mean APT signals showed excellent accordance, with one value beyond the 95% limits of agreement. The 90-percentile APT signals showed excellent accordance, all values within the 95% limits of agreement. No additive or proportional biases were observed in either analyses.
DISCUSSION Endogenous CEST imaging suffers from small saturation transfer effects since the resonance frequency of the solute protons is sufficiently close to the bulk water frequency and the CEST effects have to compete with direct saturation of the water protons.16,17 The situation is even worse in vivo due to the large background MT effects. For example, in the study for grading diffuse gliomas, it was reported that APT SI was 2.1 6 0.4% in the grade II gliomas, 3.2 6 0.9% in the grade III gliomas, and 4.1 6 1.0% in the grade IV gliomas.8 In order to sensitively detect such small signal differences, APT imaging must be consistent and reproducible.
FIGURE 8: Mean (A) and 90-percentile (B) value of the APT signal of a whole tumor ROI in the two scans in the single session without repositioning. Excellent correlations between the two scans in a session are observed in the mean APT signal (R2 5 0.96, P < 0.0001) and in the 90-percentile APT signal (R2 5 0.93, P < 0.0001).
November 2015
1351
Journal of Magnetic Resonance Imaging
FIGURE 9: Bland-Altman plot analysis of the mean (A) and 90-percentile (B) of APT signal of tumor in the two scans in the single session. The mean and 90-percentile APT signals show excellent accordance, with only one and no value beyond the 95% limits of agreement (broken line), respectively. No additive or proportional biases were observed in either analysis.
Spatial homogeneity of the saturation pulse is essential to ensure consistent APT imaging. Parallel transmission based APT imaging incorporated special RF shimming, which enables homogenized (adaptively shimmed) RF saturation pulses by adjusting the individual power of the RF amplifiers during alternation, according to the results of the RF calibration scan (B1 calibration). A previous study conducted in a pasteurized, homogeneous chicken egg-white phantom showed that gradient-echo-based APT image without RF shimming was spatially heterogeneous with 3% locational variation in APT signal in the egg-white phantom.11 On the other hand, the gradient-echo based APT image with RF shimming showed much more homogenous APT signal in the phantom with 1% locational variation. Furthermore, FSE with driven equilibrium sequence with RF shimming, which was used in this study, showed even more homogenous saturation effect with only 0.5% locational variation in APT signal, compared with the gradientecho sequence.11 Another critical point for consistent APT imaging is B0 inhomogeneity correction. Endogenous CEST imaging is extremely sensitive to magnetic field inhomogeneity.16,17 Due to the steep slope of the direct saturation curve of the water, even a small B0 field difference and a resulting shift in the Z-spectrum may cause a large change in MTRasym, where the local magnetic susceptibility effects are commonplace. In vivo, this effect may result in the occurrence of signal variation in APT images. To reduce this problem, image acquisition techniques have been performed that included many saturation frequency offsets, followed by polynomial or cubic-spline fitting and centering of the Z-spectrum in each voxel.2 This B0 correction method was applicable in APT imaging because the frequency offset of amide protons (13.5 ppm) is relatively large, eg, compared to hydroxyl protons (1 ppm). However, this correction approach is sometimes problematic when the MT effect of the tissue is large and the Z-spectrum is broad, in which the center of water frequency is difficult to determine accurately. In addi1352
tion, the conventional MT effect in the normal brain has been shown to be asymmetric.1,18 As an alternative, a B0 map obtained from the phase images of a dual-echo gradient-echo sequence was used in this study. B0 shift correction using a B0 map is a simple and robust method since this centering method is not affected by the MT effect and direct saturation of water. In this study, the mean and 90percentile of APT signal of tumor showed excellent agreement between the two sessions (between-session, B0 inhomogeneity corrected with different B0 maps), which was comparable to the agreements between the two scans in a single session (within-session, B0 inhomogeneity corrected with the same B0 map). These results indicate that B0 mapbased B0 inhomogeneity correction is accurate and reproducible. Another advantage of the use of a B0 map is that it is not necessary to obtain a full Z-spectrum and thus scan time can be considerably shortened. One scan in the second session in this study was obtained using only seven saturation frequency offsets around 63.5 ppm, which yielded almost identical APT-weighed maps to the other scan obtained from the full Z-spectrum acquisition (25 saturation frequency offsets) in spite of the shorter acquisition time (2 min 20 sec vs. 35 sec). The limitations of the study are as follows. The number of patients for each type of tumor and the total number of patients involved are small. There is no real reference standard for APT signal. Only a single observer drew an ROI; however, observer bias should have been small since the ROI was placed to include the whole tumor area. The slope in the between-session comparison deviated from 1 by 12% in mean APT signal and 16% in 90-percentile APT signal, whereas that in the within-session comparison was nearly 1. These deviations in between-session analysis can be attributed to difficulty in positioning the single slice to the exact same location and some variability in B0 shimming. The use of a single slice acquisition to reduce total scan time, since the APT imaging was included in the clinical preoperative scans for the patients. The multislice 2D and Volume 42, No. 5
Togao et al.: Reproducibility of APT Imaging 6.
Wen Z, Hu S, Huang F, et al. MR imaging of high-grade brain tumors using endogenous protein and peptide-based contrast. Neuroimage 2010;51:616–622.
7.
Togao O, Kessinger CW, Huang G, et al. Characterization of lung cancer by amide proton transfer (APT) imaging: an in-vivo study in an orthotopic mouse model. PLoS One 2013;8:e77019.
8.
Togao O, Yoshiura T, Keupp J, et al. Amide proton transfer imaging of adult diffuse gliomas: correlation with histopathological grades. Neurol Oncol 2014;16:441–448.
9.
Sagiyama K, Mashimo T, Togao O, et al. In vivo chemical exchange saturation transfer imaging allows early detection of a therapeutic response in glioblastoma. Proc Natl Acad Sci U S A 2014;111:4542– 4547.
10.
Zhou J, Tryggestad E, Wen Z, et al. Differentiation between glioma and radiation necrosis using molecular magnetic resonance imaging of endogenous proteins and peptides. Nat Med 2011;17:130–134.
11.
Keupp J, Baltes C, Harvey P, Van den Brink J. Parallel RF transmission based MRI technique for highly sensitive detection of amide proton transfer in the human brain at 3T. In: Proc 19th Annual Meeting ISMRM, Montreal; 2011.
12.
Petersen ET, Mouridsen K, Golay X. The QUASAR reproducibility study. Part II. Results from a multi-center arterial spin labeling testretest study. Neuroimage 2010;49:104–113.
13.
Louis DN, Ohgaki H, Wiestler OD, et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 2007; 114:97–109.
14.
Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86:420–428.
15.
Zhou J, Payen JF, Wilson DA, Traystman RJ, van Zijl PC. Using the amide proton signals of intracellular proteins and peptides to detect pH effects in MRI. Nat Med 2003;9:1085–1090.
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307– 310.
16.
2.
Zhou J, Lal B, Wilson DA, Laterra J, van Zijl PC. Amide proton transfer (APT) contrast for imaging of brain tumors. Magn Reson Med 2003; 50:1120–1126.
Zhou J, Blakeley JO, Hua J, et al. Practical data acquisition method for human brain tumor amide proton transfer (APT) imaging. Magn Reson Med 2008;60:842–849.
17.
3.
van Zijl PC, Jones CK, Ren J, Malloy CR, Sherry AD. MRI detection of glycogen in vivo by using chemical exchange saturation transfer imaging (glycoCEST). Proc Natl Acad Sci U S A 2007;104:4359–4364.
Kim M, Gillen J, Landman BA, Zhou J, van Zijl PC. Water saturation shift referencing (WASSR) for chemical exchange saturation transfer (CEST) experiments. Magn Reson Med 2009;61:1441–1450.
18.
4.
Ling W, Regatte RR, Navon G, Jerschow A. Assessment of glycosaminoglycan concentration in vivo by chemical exchange-dependent saturation transfer (gagCEST). Proc Natl Acad Sci U S A 2008;105:2266–2270.
Zhou J, Hong X, Zhao X, Gao JH, Yuan J. APT-weighted and NOEweighted image contrasts in glioma with different RF saturation powers based on magnetization transfer ratio asymmetry analyses. Magn Reson Med 2013;70:320–327.
5.
Salhotra A, Lal B, Laterra J, Sun PZ, van Zijl PC, Zhou J. Amide proton transfer imaging of 9L gliosarcoma and human glioblastoma xenografts. NMR Biomed 2008;21:489–497.
19.
Keupp J, Eggers H. CEST-DIXON MRI for sensitive and accurate measurement of amide proton transfer in humans at 3T. In: Proc 18th Annual Meeting ISMRM, Stockholm; 2010.
3D volume coverage acquisition with B0 inhomogeneity correction function is now under development. Possible subtle position shifts between the APT imaging and B0 map acquisition could have affected the quantitation of the APT signals. Efficient self-B0 correcting CEST methods using multiecho gradient-echo/fast spine-echo (CEST-DIXON) are expected to overcome this issue.19 In conclusion, we demonstrated a good reproducibility of the parallel transmission-based APT imaging of brain tumors confirmed by the scan–rescan test. Spatially homogenous RF shimming and accurate B0 inhomogeneity correction could have greatly contributed to the reproducibility.
Acknowledgment Contract grant sponsor: Japanese Society of Neuroradiology; Contract grant sponsor: Japanese Radiological Society; Contract grant sponsor: Philips Electronics Japan; Contract grant sponsor: Bayer Healthcare Japan; Contract grant sponsor: Fukuoka Foundation for Sound Health Cancer Research Fund; Contract grant numbers: Grant-in-Aid for Scientific Research numbers 26461827 and 22591340.
REFERENCES 1.
November 2015
1353
