Motion artifact reduction of diffusion-weighted MR imaging of the liver: use of velocity-compensated diffusion gradients combined with tetrahedral gradients

(1)

Motion artifact reduction of

diffusion‑weighted MR imaging of the liver:

use of velocity‑compensated diffusion

gradients combined with tetrahedral gradients

著者尾? 正則

著者別表示 Ozaki Masanori journal or

publication title

博士論文本文Full 学位授与番号 13301甲第4120号

学位名博士（保健学）

学位授与年月日 2014‑09‑26

URL http://hdl.handle.net/2297/41338

doi: 10.1002/jmri.23796

Creative Commons : 表示 ‑ 非営利 ‑ 改変禁止 http://creativecommons.org/licenses/by‑nc‑nd/3.0/deed.ja

(2)

Original Research

Motion Artifact Reduction of Diffusion-Weighted MRI of the Liver: Use of Velocity-Compensated Diffusion Gradients Combined With

Tetrahedral Gradients

Masanori Ozaki, MS,

^1,2

* Yusuke Inoue, MD, PhD,

³

Tosiaki Miyati, PhD, DMSc,

²

Hirohumi Hata, RT,

⁴

Sinya Mizukami, RT,

⁴

Shotaro Komi, RT,

⁴

Keiji Matsunaga, MD,

³

and Reiko Woodhams, MD, PhD

³

Purpose:To assess the effect of motion artifact reduction on the diffusion-weighted magnetic resonance imaging (DWI-MRI) of the liver, we compared velocity-compensated DWI (VC-DWI) and VC-DWI combined with tetrahedral gradients (t-VC-DWI) to conventional DWI (c-DWI) in the assessment of apparent diffusion coefficients (ADCs) of the liver.

Materials and Methods: In 12 healthy volunteers, the liver was scanned with c-DWI, VC-DWI, and t-VC-DWI sequences. The signal-to-noise ratio (SNR) and ADC of the liver parenchyma were measured and compared among sequences.

Results: The image quality was visually better for t-VC- DWI than for the others. The SNR for t-VC-DWI was significantly higher than that for VC-DWI (P < 0.05) and comparable to that for c-DWI. ADCs in both hepatic lobes were significantly lower for t-VC-DWI than for c-DWI (P<

0.01). ADC in the left lobe was significantly lower for VC- DWI than for c-DWI (P <0.01). Although ADC in the left lobe was significantly higher for c-DWI (P<0.01), no significant differences in ADCs were found between the right and left lobes for VC-DWI and t-VC-DWI.

Conclusion:The use of a t-VC-DWI sequence enables us to correct ADCs of the liver for artificial elevation due to cardiac motion, with preserved SNR.

Key Words: magnetic resonance imaging; diffusion- weighted imaging; apparent diffusion coefficient; liver;

cardiac motion; velocity compensation J. Magn. Reson. Imaging 2013;37:172–178.

VC 2012 Wiley Periodicals, Inc.

DIFFUSION-WEIGHTED (DW) magnetic resonance imaging (MRI) with the calculation of apparent diffusion coefficient (ADC) values is a valuable MRI technique for the detection and characterization of liver lesions (1–5). However, previous reports demonstrate that DW MRI (DWI) of the liver is affected by cardiac and respiratory motion (6–12).

DWIs of the liver may be performed during a breath hold to freeze respiratory motion, or during free breathing with multiple signal acquisitions or with respiratory triggering to reduce the effects of respiratory motion. The use of free breathing with multiple signal acquisitions offers images with high signal-to-noise ratio (SNR). High-quality DW images of the liver can be obtained using a free-breathing technique because cyclical respiration is a coherent motion that does not cause additional attenuation of the signal from the liver (13). However, free breathing with multiple signal acquisitions impairs the reliability and reproducibility of ADC measurements for characterizing liver lesions, because data for different axes of diffusion gradients are acquired at different respiratory phases (6). There- fore, it is useful to perform DWI of the liver in combination with respiratory triggering, although the use of respiratory-triggered acquisition increases the acquisition time.

Furthermore, a typical pulse sequence for DWI adds a bipolar gradient in three orthogonal directions, after excitation and before readout. These pulsed gradients will induce a phase shift due to cardiac motion as well as cause attenuation of the echo signal due to diffusion of the spins (7–9). Such a phase shift can cause signal loss on DWI and artificial elevation of ADC values in the liver (7,8), and the problems are

1School of Allied Health Sciences, Kitasato University, Kanagawa, Japan.

2Graduate School of Medical Science, Kanazawa University, Kanagawa, Japan.

3Department of Diagnostic Radiology, Kitasato University School of Medicine, Kanagawa, Japan.

4Department of Radiology, Kitasato University Hospital, Kanagawa, Japan.

Contract grant sponsor: Kitasato University School of Allied Health Sciences; Contract grant number: Grant-in-Aid for Research Project, No. 2011-1040.

*Address reprint requests to: M.O., School of Allied Health Sciences, Kitasato University, 1-15-1 Kitasato, Minami-ku, Sagamihara, Kana- gawa, Japan, 252-0373. E-mail: m.ozaki@kitasato-u.ac.jp

Received November 25, 2011; Accepted July 31, 2012.

DOI 10.1002/jmri.23796

View this article online at wileyonlinelibrary.com.

JOURNAL OF MAGNETIC RESONANCE IMAGING 37:172–178 (2013)

CME

VC 2012 Wiley Periodicals, Inc. 172

(3)

larger at higher b-values. One method of minimizing such artifacts is the use of cardiac triggering at image acquisition (10). When cardiac triggering is employed, higher reliability and reproducibility of ADC measurements in the liver may be obtained. However, the use of cardiac triggering, as well as respiratory triggering, increases the acquisition time. Combining cardiac- and respiratory-triggered acquisition further increases the acquisition time and is not suitable for routine clinical use.

The use of a velocity-compensated (VC)-DWI sequence is another method to depress artifacts due to cardiac motion (Fig. 1) (14–16). VC diffusion gradients use dual bipolar gradients to remove all phase sensitivity to first-moment velocity motion during the diffusion-weighting period without cardiac triggering.

One limitation of VC-DWI is that, for a given b-value, the duration of the VC diffusion gradient pulses becomes 1.6 times longer than that of the conventional Stejskal-Tanner (ST) diffusion gradients pulses, which leads to a reduction in the SNR (13). The VC- DWI sequence is appropriate for neonatal brain imaging because of the long T₂ values of the brain (17).

However, the short T2values of the liver make it diffi- cult to acquire images of sufficient SNR using a VC- DWI sequence in clinical settings.

In a conventional DWI sequence, orthogonal x, y, and z diffusion gradient pulses are separately applied, and isotropic DW images are obtained by using three orthogonal images. This method is referred to as the orthogonal-gradients method (Fig. 2a). Orthogonal x, y, and z diffusion gradients pulses may be simultane- ously applied, which shortens the duration of diffusion gradient pulses without reducing the b-value, resulting in a higher SNR. This method can generate four different vectors arranged in tetrahedral directions, called tetrahedral gradients (Fig. 2b). Isotropic DW images are obtained by using four different tetrahedral vector images (18). Recent reports have shown

that the tetrahedral gradients provide a higher SNR than orthogonal gradients (19,20). Therefore, incorpo- rating tetrahedral gradients into a VC-DWI sequence can be expected to improve SNR.

In the present study, we compared a VC-DWI sequence and a VC-DWI sequence combined with tetrahedral gradients (t-VC-DWI) to conventional DWI (c-DWI) in the assessment of liver ADC values.

MATERIALS AND METHODS Subjects

This study was approved by the local Institutional Review Board and written informed consent was obtained from all participants. Twelve healthy adult volunteers (eight men and four women; mean age, 30.7 years; age range, 22–38 years) with no prior history or findings related to liver disease at the time of the study prospectively underwent DWI of the liver. Exclusion criteria were general contraindications to MRI, such as implanted pacemaker and claustrophobia.

Imaging Procedures

All examinations were performed on a clinical MR scanner (Signa HDxt optima edition 1.5T, GE Health- care, Milwaukee, WI). The system provides a maxi- mum gradient strength of 33 mT/m with a peak slew rate of 120 mT/m/msec. The coils used here were a 12-element body phased-array coil. DWI of the liver was performed in each subject using all three sequences: c-DWI, VC-DWI, and t-VC-DWI. The parameters of single-shot spin-echo echo planar imaging were as follows: axial-plane image acquisition, repetition time (TR) 12,610–20,000 msec (with respiratory triggering), echo time (TE) 76.3 msec (c-DWI), 121.6 msec (VC- DWI), or 95.5 msec (t-VC-DWI), slice thickness/slice gap 10/0 mm, number of slices 15, receive bandwidth 6250 kHz, b-values 0 and 1000 s/mm², field of view (FOV) 400 mm, matrix 128, signal averaging 4 (for the orthogonal gradients) or 3 (for the tetrahedral gradients), and spectral spatial radiofrequency fat suppression. Orthogonal gradients were used for c-DWI Figure 1. Pulse sequence charts of conventional diffusion

sequence (a) and velocity-compensated diffusion sequence (b). The radiofrequency (RF) pulses (excite and refocus), diffusion gradients (Gdiff) and echo planar imaging (EPI) readout are shown; other sequence elements are omitted for simplic- ity. Dual bipolar diffusion gradients enable velocity compensation (b).

Figure 2. Schematic presentation of orthogonal gradients (a), and tetrahedral gradients (b). The x, y, and z directions indicate the left, anterior, and caudal directions, respectively, for a patient placed in the magnet head-first and in the supine position. Since the tetrahedral gradients have strength equal to the square root of 3 times the orthogonal input components, this technique shortens the duration of the diffusion gradient pulse (18).

(4)

and VC-DWI. Mean nominal acquisition time was 231 seconds (range, 187–306 sec) for c-DWI, 239 seconds (range, 168–306 sec) for VC-DWI, and 232 seconds (range, 183–306 sec) for t-VC-DWI.

Visual Assessment

Images obtained with c-DWI, VC-DWI, and t-VC-DWI sequences were visually evaluated by two board-certified radiologists in a random and blinded manner. Se- verity of signal loss in the right hepatic lobe, severity of signal loss in the left hepatic lobe, and degree of noise were assessed using a three-point grading scale (1 ¼ severe, 2 ¼ mild, 3 ¼ negligible), and overall image quality was assessed using another three-point grading scale (1¼ poor, 2 ¼ fair, 3¼ good). Discrep- ancy was resolved by the third board-certified radiol- ogist. The visual assessment was performed on the monitor of Advantage Workstation (GE Healthcare), and the observers were allowed to adjust the window level and width of the DW images.

Quantitative Analysis

ADC maps were generated from isotropic DW images and T₂-weighted images with a b-value of 0 s/mm² using the following equation: ADC¼ - ln (S(b)/S(0))/b, where b is the b-value and S_(b) and S₍₀₎ are the signal intensities on images with b-values of 1000 and 0 s/

mm², respectively.

For SNR and ADC measurements, regions of inter- est (ROIs) were placed in the liver parenchyma, excluding visible vascular and biliary structures, on the slice including the umbilical portion. SNRs were calculated by the method proposed by Steckner (21) (Appendix).

ROIs for measuring SNR were set to the smallest possible size in order to minimize the effect of spa- tially variant noise caused by parallel imaging recon- struction techniques. Three ROIs were set in the right hepatic lobe (anterior, lateral, and posterior portions) (Fig. 3a), and the SNR was calculated for each ROI.

Because the signal in the left hepatic lobe was Figure 3. Examples of ROIs placed in the liver parenchyma for calculating SNRs (a) and ADCs (b). ROIs for SNR calculation were set only in the right hepatic lobe. Red and yellow circles in panel b represent ROIs for the right and left hepatic lobes, respectively.

Figure 4. DW images (a–c) and ADC maps (d–f) acquired with c-DWI (a,d), VC-DWI (b,e), and t-VC-DWI (c,f) sequences.

(5)

affected by artifacts, SNR was not computed for the left lobe. For ADC measurement, three circular ROIs (49 pixels) were placed for each of the right and left hepatic lobes (Fig. 3b). The ADC value for each lobe was defined as the average of the ADC values determined from the three ROIs. All image analyses were performed with MatLab (MathWorks, Natick, MA).

Statistical Analysis

ADC values were compared between the right and left hepatic lobes using a paired t-test, and ADC values and SNRs were compared among c-DWI, VC-DWI, and t-VC-DWI using one-way repeated measures analysis of variance (ANOVA) and the post-hoc Tukey test, because the data fitted a normal distribution (Kolmo- gorov–Smirnov test). Visual grades were compared among c-DWI, VC-DWI, and t-VC-DWI using the Kruskal–Wallis test and the post-hoc Steel–Dwass test, and visual grades regarding the severity of signal loss were compared between the right and left hepatic lobes using the Wilcoxon signed-rank test, because the data did not fit a normal distribution.

Differences were considered significant when P <

0.05. All statistical analyses were performed on a personal computer with JMP v. 9.0 (SAS Institute, Cary, NC).

RESULTS

DW image sets using the three sequences were suc- cessfully acquired in all 12 subjects with no need for retry (Fig. 4). Table 1 presents results of visual

assessment regarding the severity of signal loss in the right hepatic lobe, severity of signal loss in the left hepatic lobe, degree of noise, and overall image quality.

All four visual grades were significantly better for t-VC-DWI than for c-DWI and for VC-DWI, although no significant differences were found between c-DWI and VC-DWI. Table 2 shows the results of comparison of visual assessments regarding the severities of signal loss between the right and left hepatic lobes.

Visual grades regarding the severity of signal loss in the left hepatic lobe were significantly worse than those in the right hepatic lobe for VC-DWI and t-VC-DWI, although no significant differences were found for c-DWI.

Figure 5 shows SNRs for each pulse sequence. The SNR was lower for VC-DWI than for c-DWI, and statis- tically significant differences were found in the anterior and lateral portions of the right hepatic lobe. No significant differences in SNRs in any of the three portions were found between c-DWI and t-VC-DWI. The SNRs in all the three portions were significantly higher for t-VC-DWI than for VC-DWI; the ratio of mean SNR for t-VC-DWI to that for VC-DWI was 2.33 in the anterior portion, 1.87 in the lateral portion, and 1.94 in the posterior portion.

Figure 6 shows ADC values for each pulse sequence.

ADC values in both hepatic lobes were significantly lower for t-VC-DWI than for c-DWI and for VC-DWI.

ADC values in the left hepatic lobe were significantly lower for VC-DWI than for c-DWI. Table 3 shows the results of comparison between ADC values in the right and left hepatic lobes. Although ADC values in the left hepatic lobe were significantly higher than that in the right hepatic lobe for c-DWI, no significant differences in ADC values were found between the right and left hepatic lobes for VC-DWI and t-VC-DWI.

DISCUSSION

DWI is a useful technique for the detection of liver lesions (1–5). Furthermore, the ADC value may be helpful for characterizing liver lesions, quantifying liver fibrosis, and predicting or assessing response to chemotherapy (22,23). Typical DWI sequences are intrinsically sensitive to diffusion. However, they are generally also highly sensitive to other kinds of motion such as bulk motion, which may produce artifacts.

Table 1

Comparison of Visual Grades Among c-DWI, VC-DWI, and t-VC-DWI

Grades

c-DWI VC-DWI t-VC-DWI

Severity of signal loss in the right hepatic lobe 1.5060.52 2.0060.60 2.9260.29**^,††

Severity of signal loss in the left hepatic lobe 1.1760.39 1.5060.67 2.2560.62*^,†††

Degree of noise 2.0860.51 1.6760.49 2.9160.29***^,†††

Overall image quality 1.5060.52 1.6760.49 2.5060.52*^,††

Values are presented as means6SD.

*P<0.05 compared with c-DWI.

**P<0.01 compared with c-DWI.

***P<0.001 compared with c-DWI.

yyP<0.01 compared with VC-DWI.

yyyP<0.001 compared with VC-DWI.

Table 2

Comparison of Visual Grades for the Severity of Signal Loss Between in the Right and Left Hepatic Lobes

Sequence

Grades

Pvalue Severity of

signal loss in the right hepatic lobe

Severity of signal loss in the left hepatic lobe

c-DWI 1.5060.52 1.1760.39 0.125

VC-DWI 2.0060.60 1.5060.67 0.031 t-VC-DWI 2.9260.29 2.2560.62 0.008 Values are presented as means6SD.

(6)

Such artifacts may influence the accuracy of ADC values (7–9).

First, the present study showed that ADC values were significantly higher in the left hepatic lobe than in the right hepatic lobe for c-DWI. Nasu et al (7) demonstrated signal loss in the left hepatic lobe on DWI and ascribed it to artifacts due to cardiac motion. The right-to-left difference in ADC observed in our study appears to be attributable to artificial elevation of ADC values of the left hepatic lobe due to cardiac motion. In contrast, no significant right-to-left differences in ADC values were observed for VC-DWI and t-VC-DWI, which indicates that VC-DWI and t-VC- DWI sequences allow correction for artificial elevation of ADC values due to cardiac motion.

Moreover, ADC values in the right hepatic lobe were significantly higher for c-DWI than for t-VC-DWI. This result suggests that signal loss on DWI and artificial elevation of ADC values also occurs in the right hepatic lobe, albeit to a lesser degree than in the left hepatic lobe, and that the artificial elevation was reduced by the use of a t-VC-DWI sequence. Kwee et al (8) indicated that signal loss due to cardiac motion also occurs in the right hepatic lobe. Nasu et al (12) demonstrated signal loss in the right hepatic lobe on respiratory-triggered DWI. They ascribed this phenomenon to artifacts originating from respiratory motion and called them ‘‘pseudo-anisotropy artifacts.’’

In our study, all DW images were acquired in combination with respiratory triggering. The artificial elevation of ADC values in the right hepatic lobe for c-DWI may have occurred due to cardiac motion and pseudo-anisotropy artifacts.

The SNR for the t-VC-DWI sequence was higher than for VC-DWI. For evaluating SNR, the sum of the axes of diffusion gradients and signal averaging was set to be equal among different pulse sequences. The use of VC gradients decreases the SNR of the liver parenchyma, despite a reduction in motion artifacts, because the resulting prolongation of effective TE leads to T₂ suppression in the liver with relatively short T2 relaxation time. In contrast, the use of tetrahedral gradients shortens the duration of diffusion gradients with the diffusion b-value kept constant, and thus provides a shorter TE, which would be expected to increase the SNR (18–20). Actually, the SNR was improved by approximately two times for t- VC-DWI compared to VC-DWI, which could be explained by the different effective TEs: 95.5 msec with t-VC-DWI sequence and 121.6 msec with the VC- DWI sequence. Although the effective TE was longer for t-VC-DWI than for c-DWI, no significant differences in SNRs were found between c-DWI and t-VC- DWI. This observation may be attributable to signal loss due to bulk motion (eg, cardiac motion) in images acquired with a c-DWI sequence, and compensation Figure 5. SNRs using each pulse sequence in the anterior (a), medial-lateral (b), and posterior (c) portions of the right hepatic lobe. Plots indicate data for individual SNRs. Mean and standard deviations are also presented. *P<0.05; **P<0.01.

Figure 6. ADC values using each pulse sequence in the left (a) and right (b) hepatic lobes. Plots indicate data for individual ADC values. Mean and standard deviations are also presented. *P<0.05; **P<0.01.

(7)

of signal loss in images acquired with a t-VC-DWI sequence. Furthermore, higher ADC values for VC- DWI than for t-VC-DWI may be responsible to the severe SNR for VC-DWI. The severe SNR for a VC-DWI sequence would prevent its clinical application in liver imaging, although imaging with a lower b-value may provide an acceptable SNR. In contrast, the t-VC-DWI sequence has the capability of providing high-quality DW images during a reasonable acquisition time even with a b-value of 1000 s/mm²and appears to be suitable for practical use.

Visual grades were significantly better for t-VC-DWI than for c-DWI and for VC-DWI. These results appear to support the utility of a t-VC-DWI sequence; however, some discrepancies were noted between quantitative analysis and visual assessments. Although quantitative evaluation of the SNR did not indicate significant differences between c-DWI and t-VC-DWI, visual evaluation of degree of noise gave significantly better results for t-VC-DWI. Quantitative evaluation was performed for the right hepatic lobe, and visual evaluation was performed considering the entire liver.

This difference may be responsible for the difference in results. Additionally, although the use of a t-VC- DWI sequence eliminated the right-to-left differences in ADC values, visual signal loss was severer for the left hepatic lobe. ROIs for quantitative analysis were placed only on the slice including the umbilical portion. The signal loss tended to be reduced on more caudal slices mainly delineating the right hepatic lobe, which may have contributed to the assignment of better visual grades to the right hepatic lobe. The right-to-left differences in visual signal loss suggest the need for further improvement. The use of the tracking only navigator echo technique (24), instead of respiratory triggering, provides sharp DW images in a shorter acquisition time. The combination of such a technique appears to be worth investigating.

We used b-values of 0 and 1000 s/mm² in this study because recent studies indicated the utility of DWI with this combination for the characterization of liver tumors (26,27). A b-value of 1000 s/mm²is relatively high for DWI of the liver (4), which should have decreased the SNR. It should also be noted that the fast diffusion fraction related to capillary perfusion and the slow diffusion fraction reflecting tissue diffusion were not separated in calculating ADC values using this combination of b-values. The slice thickness was set at 10 mm in this study to compare DW images with different sequences. In clinical practice, the acquisition of thinner slices is recommended.

This study had several limitations. First, t-VC-DWI has not been proven yet to provide accurate ADC values, because of the lack of standard ADC values. How- ever, t-VC-DWI provided approximately equivalent ADC values between the right and left hepatic lobes and enabled the assessment of the left lobe, which should be beneficial for clinical application. Second, only a small number of healthy volunteers were included in this study; future patient studies are necessary to determine the clinical utility of a t-VC-DWI sequence in detecting and characterizing hepatic lesions.

In conclusion, this study demonstrated that the t- VC-DWI sequence is able to compensate for artificial elevation of ADC values in the liver due to cardiac motion with maintenance of an acceptable SNR. The use of a t-VC-DWI sequence enables the acquisition of high-quality DW images even at a high b-values and has the potential to contribute to clinical liver imaging.

APPENDIX

We calculated SNRs by the method proposed by Steckner (21). In this method, both the signal intensity and noise are determined using a given ROI.

Noise is estimated excluding the contribution of artifacts such as the ringing artifact and ghost artifact to the variability of the signal intensity.

The standard deviation (SD) was calculated using the following equation:

SD¼ Xⁿ²

i¼1

ðXiþ1XiÞ² 2ðn1Þ

" #1=2

(1)

where n is the number of pixels evaluated and X_iand Xiþ1represent signal intensities of successive pixels.

In addition to the underlying noise variance of in- terest (s), additional variance due to issues found exclusively in the read (s_r) or phase (s_p) direction, including the ringing artifact and ghost artifact, was considered as the source of the standard deviation.

The directionally unique standard deviations are determined as follows:

SDr ¼ sqrtðs²þs²_rÞ (2) SDp ¼ sqrtðs²þs²_pÞ (3) SDd ¼ sqrtðs²þs²_rþs²_pÞ (4) where SD_r, SD_p, and SD_d are standard deviations in the read, phase, and diagonal directions, respectively.

The SDr, SDp, and SDd were calculated using the acquired images and Eq. [1], and the value of s was computed using the SDr, SDp, SDd, and Eqs. [2–4].

The SNR was calculated using the following equation:

SNR ¼ SI s=p

ð 2Þ (5)

where SI is the mean signal intensity in the ROI.

Table 3

ADC Values in the Right and Left Hepatic Lobes

Sequence

ADC (10³mm²/s)

Pvalue Right hepatic

lobe

Left hepatic lobe

c-DWI 1.7560.38 2.1060.37 0.006

VC-DWI 1.6660.20 1.7260.17 0.359 t-VC-DWI 1.4160.11 1.4660.12 0.101 Values are presented as means6SD.

(8)

REFERENCES

1. Low RN. Abdominal MRI advances in the detection of liver tumours and characterisation. Lancet Oncol 2007;8:525–535.

2. Naganawa S,Kawai H,Fukatsu H, et al. Diffusion-weighted imaging of the liver: technical challenges and prospects for the future.

Magn Reson Med Sci 2005;4:175–186.

3. Gourtsoyianni S, Papanikolaou N, Yarmenitis S, Maris T, Karan- tanas A, Gourtsoyiannis N. Respiratory gated diffusion-weighted imaging of the liver: value of apparent diffusion coefficient measurements in the differentiation between most commonly encoun- tered benign and malignant focal liver lesions. Eur Radiol 2008;

18:486–492.

4. Bruegel M, Holzapfel K, Gaa J, et al. Characterization of focal liver lesions by ADC measurements using a respiratory triggered diffusion-weighted single-shot echo-planar MR imaging technique. Eur Radiol 2008;18:477–485.

5. Demir OI, Obuz F, Sagol O, Dicle O. Contribution of diffusion weighted MRI to the differential diagnosis of hepatic masses.

Diagn Interv Radiol 2007;13:81–86.

6. Nasu K, Kuroki Y, Sekiguchi R, Nawano S. The effect of simultaneous use of respiratory triggering in diffusion-weighted imaging of the liver. Magn Reson Med Sci 2006;5:129–136.

7. Nasu K, Kuroki Y, Nawano S, et al. Hepatic metastases: diffusion-weighted sensitivity encoding versus SPIO-enhanced MR Imaging. Radiology 2006;239:122–130.

8. Kwee TC, Takahara T, Niwa T, et al. Influence of cardiac motion on diffusion-weighted magnetic resonance imaging of the liver.

MAGMA 2009;22:319–325.

9. Chenevert TL, Pipe JG. Effect of bulk tissue motion on quantitative perfusion and diffusion magnetic resonance imaging. Magn Reson Med 1991;19:261–265.

10. Murtz P, Flacke S, Traber F, van den Brink JS, Gieseke J, Schild HH. Abdomen: diffusion-weighted MR imaging with pulse triggered single-shot sequences. Radiology 2002;224:258–264.

11. Kwee TC, Takahara T, Koh DM, Nievelstein RA, Luijten PR.

Comparison and reproducibility of ADC measurements in breath- hold, respiratory triggered, and free-breathing diffusion-weighted MR imaging of the liver. J Magn Reson Imaging 2008;28:

1141–1148.

12. Nasu K, Kuroki Y, Fujii H, Minami M. Hepatic pseudo-anisotropy:

a specific artifact in hepatic diffusion-weighted images obtained with respiratory triggering. MAGMA 2007;20:205–211.

13. Kwee TC, Takahara T, Ochiai R, Nievelstein RA, Luijten PR. Diffu- sion-weighted wholebody imaging with background body signal suppression (DWIBS): features and potential applications in on- cology. Eur Radiol 2008;18:1937–1952.

14. Johnson GA, Maki JH. In vivo measurement of proton diffusion in the presence of coherent motion. Invest Radiol 1991;26:540–545.

15. Clark CA, Barker GJ, Tofts PS. Improved reduction of motion artifacts in diffusion imaging using navigator echoes and velocity compensation. J Magn Reson 2000;142:358–363.

16. Ogino T, Horie T, Muro I, Takahara T, Van Cauteren M. Respira- tory motion artefact reduction on abdominal DWI: simultaneous use of dual bipolar diffusion gradient and navigator slice tracking. In: Proc 17th Annual Meeting ISMRM, Honolulu; 2009 (abstract 2009).

17. Winter JD, Gelman N, Picot PA, et al. Reduction of motion artifacts in segmented EPI with velocity compensation diffusion gradients, for application to neonatal imaging at 3.0T. In:

Proc 13th Annual Meeting ISMRM, Toronto; 2003 (abstract 2129).

18. Conturo TE, McKinstry RC, Akbudak E, Robinson BH. Encoding of anisotropic diffusion with tetrahedral gradients: a general mathematical diffusion formalism and experimental results.

Magn Reson Med 1996;35:399–412.

19. Hori M, Kim T, Murakami T, et al. Isotropic diffusion-weighted MR imaging with tetrahedral gradients in the upper abdomen.

Magn Reson Med Sci 2006;5:201–206.

20. Ueguchi T, Yamada S, Mihara N, Koyama Y, Sumikawa H, Tomiyama N. Breast diffusion-weighted MRI: comparison of tetrahedral versus orthogonal diffusion sensitization for detection and localization of mass lesions. J Magn Reson Imaging 2011;33:

1375–1381.

21. Steckner MC. A simple method for estimating the noise level in a signal region of an MR image. Med Phys 2010;37:5072–5079.

22. Taouli B, Tolia AJ, Losada M, et al. Diffusion-weighted MRI for quantification of liver fibrosis: preliminary experience. AJR Am J Roentgenol 2007;189:799–806.

23. Girometti R, Furlan A, Bazzocchi M, et al. Diffusion-weighted MRI in evaluating liver fibrosis: a feasibility study in cirrhotic patients. Radiol Med (Torino) 2007;112:394–408.

24. Takahara T, Kwee TC, Van Leeuwen MS, et al. Diffusion- weighted magnetic resonance imaging of the liver using tracking only navigator echo: feasibility study. Invest Radiol 2010;45:

57–63.

25. Ivancevic MK, Kwee TC, Takahara T, et al. Diffusion-weighted MR imaging of the liver at 3.0 Tesla using tracking only navigator echo (TRON): a feasibility study. J Magn Reson Imaging 2009;30:

1027–1033.

26. Erturk SM, Ichikawa T, Sano K, Motosugi U, Sou H, Araki T.

Diffusion-weighted magnetic resonance imaging for characterization of focal liver masses: impact of parallel imaging (SENSE) and b value. J Comput Assist Tomogr 2008;32:

865–871.

27. Demir OI, Obuz F, Sagol O, Dicle O. Contribution of diffusion- weighted MRI to the differential diagnosis of hepatic masses.

Diagn Interv Radiol 2007;13:81–86.