Original Research

Comparison of High Spatial Resolution RespiratoryTriggered Inversion Recovery-Prepared SpoiledGradient Echo Sequence With Standard BreathholdT1 Sequence MRI of the Liver Using Gadoxetic Acid

Phillip V.P. Tran, MBBS, FRANZCR, and Kartik S. Jhaveri, MD*

Purpose: To assess if a high resolution respiratory trig-gered inversion recovery prepared GRE sequence (RT)improved image quality and detection of lesions comparedwith breathhold GRE T1 weighted MR sequence (BH) inthe hepatobiliary uptake phase of MR of the liver usinggadoxetic acid (Gd-EOB-DTPA).

Materials and Methods: Thirty-eight consecutive patientsfrom July 2009 to September 2010 who had undergoneGd-EOB-DTPA enhanced liver exams were retrospectivelyidentified. Qualitative assessment performed on referencelesions and background liver by two independent readers.Quantitative assessment performed by one reader.

Results: Liver parenchyma signal-to-noise ratio for BHwas 90.3 6 23.9 (mean 6 SD) and RT, 106.1 6 40.4 (P ¼0.119). For BH, 320 lesions were detected compared with257 for RT. Lesion to liver contrast was significantly bet-ter on RT sequences (0.26 6 0.24; mean 6 SD) comparedwith BH sequence (0.21 6 0.20; P ¼ 0.044).

Fifty-seven reference lesions assessed. Both reviewersrated BH better for lesion margin and hepatic vesselsharpness. BH was rated with less artifact (P < 0.05).Lesion to liver contrast on BH was significantly better forone reviewer.

Conclusion: BH sequence had better overall image qual-ity than RT in several quantitative and qualitative factorsincluding number of lesions detected and level of artifact.

Key Words: Gadolinium DTPA; magnetic resonance imag-ing; liver/metabolism; liver neoplasms; biliary tract/metabolismJ. Magn. Reson. Imaging 2013;37:700–706.VC 2013 Wiley Periodicals, Inc.

THE DEVELOPMENT OF liver-specific MR contrastagents has improved detection of liver lesions andaids in the characterization of malignant and benignlesions (1–3). Gradient echo (GRE) breathhold (BH)T1-weighted sequences are frequently used in stand-ard liver MR protocols due to improved image qualityover non-breathhold spin echo sequences and fasteracquisition times (4). Acquisition in the hepatobiliaryphase with Gd-EOB-DTPA is not critically time sensi-tive as the dynamic phase. Thus a novel respiratorytriggered (RT), inversion recovery (IR), high resolutionT1 GRE sequence was reported to provide betterimage quality than BH at the expense of longer acqui-sition times (5). Even with the fast acquisition timesusing GRE, some patients are uncooperative or stillunable to breathhold for sufficient periods to allow fordiagnostic images (6). Thus, there is a role for use of arobust respiratory triggered sequence in clinical MRimaging of the liver. The other benefits of the novelsequence have been previously reported to be higherspatial resolution, improved signal to noise ratio(SNR) and greater anatomical coverage.

The purpose of this study was to assess if the highspatial resolution RT IR GRE sequence improvedimage quality and detection of lesions compared withBH GRE based on a wider inclusion criteria comparedwith previously published data.

MATERIAL AND METHODS

Subjects

Research and ethics board approval was obtained forthis retrospective analysis and the need for patientconsent was waived. A search of the RIS databaseidentified 114 consecutive patients with the termsMRI, liver, gadoxetic acid, and Primovist from January2009 to September 2010. Thirty-eight consecutivepatients were identified with a GD-EOB-DTPAenhanced liver MR exam, wherein both BH and RTsequences were available for the hepatobiliary phase,for assessment of focal liver lesions. Seventy sixpatients were excluded due to lack of gadoxeticacid or a respiratory triggered sequence. The study

Department of Medical Imaging, University of Toronto, UniversityHealth Network and Mount Sinai Hospital, Toronto, ON, Canada.

*Address reprint requests to: K.S.J., Department of Medical Imaging,University of Toronto, University Health Network and Mount SinaiHospital, 610 University Avenue, 3-957, Toronto, ON M5G 2M9,Canada. E-mail: kartik.jhaveri@uhn.ca

Received February 19, 2012; Accepted September 6, 2012.

DOI 10.1002/jmri.23864View this article online at wileyonlinelibrary.com.

JOURNAL OF MAGNETIC RESONANCE IMAGING 37:700–706 (2013)

CME

VC 2013 Wiley Periodicals, Inc. 700

population had 34 females and 3 males with a meanage of 44.26 years 6 13.81 (range, 20–77 years).

MRI Technique

Both 1.5 Tesla (T) and 3.0T systems (MagnetomAvanto and Verio, respectively, Siemens Healthcaresector, Erlangen, Germany) were used. In our institu-tion, the allocation of patients to either machine isdetermined by clerical staff based on availability.Phased array body coils were used on all patients.Gadoxetic acid (Gd-EOB-DTPA, Primovist, Bayer-Schering Pharma AG, Berlin, Germany) was adminis-tered intravenously at a dose of 0.025 mmol/kg ofbody weight. The hepatobiliary phase was acquired at20 min following bolus injection. The MR parametersused are displayed in Table 1. A standard axial BHGRE T1 weighted volume interpolated breathholdenhanced (VIBE) sequence was obtained first. A respi-ratory triggered IR GRE T1 weighted sequence (5) wasperformed after the sagittal and coronal BH VIBEsequences were acquired. Respiratory triggering wasperformed with a navigator sequence which monitors

the motion of the right hemidiaphragm. Data isacquired only when the hemidiaphragm is within theacceptance window, which was user defined at endexpiration. Inversion recovery preparation was used toimprove T1 contrast. Fat saturation was achievedwith water excitation.

Image Analysis

Qualitative

All cases were anonymized and each sequenceassigned a random case number. Two reviewers expe-rienced in abdominal MRI (reviewer 1, 10 years; andreviewer 2, 2 years) independently reviewed the cases.The reference lesions and background liver wereassessed using the grading scale in Table 2.

Quantitative

Reviewer 2 performed all quantitative analysis. Meas-urements performed on eFilm workstations (MergeHealthcare, Chicago, IL, USA). Liver signal to noise(SNR) ratio was calculated according to equation 1 (5,7).

SNRðliverÞ ¼ SIðtissueÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2

4�pSTDðairÞ

q ½1�

SI(tissue) was calculated by averaging the meanvalue of four equal region of interest (ROI) measure-ments across all liver segments at the level of theporta hepatis. STD (air) was calculated by averagingfour standard deviation of ROI measurements placedin the corners of the image in air.

Lesion to liver contrast (LLC) was calculated accord-ing to equation 2 (5).

LLC ¼ SIðlesionÞ � SIðliverÞSIðlesionÞ þ SIðliverÞ ½2�

Equal sized ROIs were used for both sequences. SI(liver) was calculated from the average of four meas-urements taken in liver parenchyma adjacent to thelesion. Lesions were classified according to high, low,or iso-intensity. For multiple lesions, the largest

Table 1

MR Parameters for RT and BH at 1.5T and 3.0T

MR

parameters 1.5T 3.0T

BH RT BH RT

TR (ms) 3.5 1800 3.5–3.89 1500–1900

TE (ms) 1.22–1.38 5.54–5.61 1.47 1.80–1.95

TI (ms) NA 800 NA 800

Flip angle

(degrees)

12 15 12 15

Matrix rows

(pixels)

230–320 312 230–260 312–336

Matrix columns

(pixels)

256–320 384 320 384

Bandwidth

(Hz/ pixel)

490 200 490 405

Slice thickness

(mm)

4.0 1.5–3.0 4.0 1.5–3.0

NA ¼ Not applicable.

Table 2

Qualitative Grading Scale Used for Assessment of Liver and Reference Hepatic Lesions

Score

1 2 3 4 5

Lesion

Lesion sharpness Sharp boundary Minimal blurring Ill- defined Lesion not visible

Lesion intensity Hyperintense Isointense Hypointense

Lesion to liver

contrast

High Moderate Low

Background liver

Liver enhancement Optimal Average Suboptimal

Hepatic vessel

sharpness

Sharp boundary Minimal blurring Moderate blurring Severe blurring Vessel not visible

Level of artifact No artifact Minimal artifact Moderate artifact Severe artifact

CBD visualized 100% 75% 50% 25% Not visualized

CBD sharpness Sharp boundary Minimal blurring Moderate blurring Severe blurring CBD not visible

CBD ¼ common bile duct.

RT Versus BH GRE MRI Liver Using Gd-EOB-DTPA 701

lesion was selected as a reference. For mixed intensitylesions, the predominant intensity was used to clas-sify the lesion. SI (lesion) was measured from themost uniform area.

Assessment of the number of lesions detected wasperformed by side-by-side analysis of the two sequen-ces. When there was a discrepancy between thesequences, additional sequences (including diffusionweighted imaging, T1 in and out of phase, T2 6 fatsaturated and dynamic phase contrast) werereviewed. A lesion was considered a true finding ifpresent on one or more of the additional sequences.A lesion was considered artifactual and disregardedif not present on any additional sequences.

Statistical Analysis

Statistical analysis of the quantitative (lesion to livercontrast, SNR, lesion number) and qualitative factorswas performed using the Wilcoxon signed ranked test.A P value of <0.05 indicated statistically significantdifferences. Inter-reader reliability was calculatedwith Cohen’s Kappa co-efficient (8). Kappa values of0–0.20 were considered as slight agreement, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 assubstantial, and 0.81–1.00 as almost perfect agree-ment. Analysis was performed with SPSS software(SPSS, version 19, IBM, Chicago, IL).

RESULTS

Qualitative Analysis

Both reviewers ranked BH sequence significantly bet-ter for lesion sharpness (reviewer 1; 1.72 6 0.96 forBH compared with 2.20 6 1.21 for RT, P ¼ 0.004 and

reviewer 2; 1.86 6 0.92 for BH compared with 2.21 6

1.16 for RT, P ¼ 0.020), CBD sharpness (reviewer 1;1.47 6 0.73 for BH compared with 2.19 6 1.04 forRT, P < 0.001 and reviewer 2; 1.95 6 0.74 for BHcompared with 2.49 6 0.86 for RT, P ¼ 0.002), andlevel of artifact (reviewer 1; 2.04 6 0.65 for BH com-pared with 2.84 6 0.92 for RT, P < 0.001 and reviewer2; 1.88 6 0.47 for BH compared with 2.96 6 0.68 forRT, P < 0.001) compared with RT sequence (Table 3).The RT sequence demonstrated significantly higherscores for background enhancement of the liver forreviewer 1 (1.59 6 0.68 compared with 1.91 6 0.63;P ¼ 0.001) but not reviewer 2. The BH sequence dem-onstrated significantly better lesion to liver contrastfor reviewer 2 (1.49 6 0.72 compared to 1.69 6 0.79;P ¼ 0.001) but not for reviewer 1. No statistical differ-ence was found between sequences for lesion intensitycompared with background liver, sharpness of hepaticvessels, and percentage of CBD visualized for bothreviewers. There was moderate agreement (8) betweenthe two readers with k ¼ 0.537 (Table 4).

Quantitative Analysis

The SNR of the background liver for the 38 patientsshowed no statistical significance (P ¼ 0.119) withmean BH SNR 90.3 6 23.9 (mean 6 SD) and RT SNR106.1 6 40.4.

Overall a greater number of lesions were detectedon the BH sequence than the RT sequence with 320lesions (range of lowest to highest number of lesionsper patient; 1–41) detected on BH and 257 lesions onRT (range, 1–30). When taking into account the influ-ence of lesion intensity on the number of lesionsdetected, there was a significant difference (P ¼ 0.048)

Table 3

Qualitative Analysis Comparing BH and RT Sequences for Both Reviewers*

R1 BH

Mean 6 SD

R1 RT

Mean 6 SD R1 P value

R2 BH

Mean 6 SD

R2 RT

Mean 6 SD

R2

P value

Lesion sharpness 1.72 6 0.96 2.20 6 1.21 0.004 1.86 6 0.92 2.21 6 1.16 0.020

Lesion to liver intensity 2.21 6 0.99 2.30 6 0.96 0.414 2.17 6 1.00 2.33 6 0.93 1.000

Lesion to liver contrast 1.42 6 0.72 1.40 6 0.66 0.070 1.49 6 0.72 1.69 6 0.79 0.001

Background liver enhancement 1.91 6 0.63 1.59 6 0.68 0.001 1.56 6 0.76 1.61 6 0.68 0.439

Sharpness hepatic vessel 2.05 6 0.85 2.11 6 1.01 0.740 1.84 6 0.75 2.07 6 0.98 0.215

Artifact 2.04 6 0.65 2.84 6 0.92 <0.001 1.88 6 0.47 2.96 6 0.68 <0.001

Bile duct visualized 1.16 6 0.59 1.33 6 0.85 0.200 1.37 6 0.84 1.44 6 0.80 0.672

Sharpness CBD 1.47 6 0.73 2.19 6 1.04 <0.001 1.95 6 0.74 2.49 6 0.85 0.002

*Lower values equates to better image quality. R1 ¼ Reviewer 1; R2 ¼ Reviewer 2; BH ¼ breathhold; RT ¼ respiratory triggered.

Table 4

Inter-rater Reliability for Qualitative Analysis Between Reviewer 1 and Reviewer 2

Reviewer 2

Ratings 0 1 2 3 4 5 Total

Reviewer 1 0 32 0 1 4 1 0 38

1 0 272 115 14 1 0 402

2 0 68 144 26 4 0 242

3 0 6 41 122 4 0 173

4 0 1 1 9 43 0 54

5 0 0 1 0 0 2 3

Total 32 347 303 175 53 2 912

*Kappa coefficient ¼ 0.537.

702 Tran and Jhaveri

in number of lesions detected on BH (mean 6.65 6 SD8.28) compared with RT (mean 5.71 6 SD 7.66) forhypointense lesions (N ¼ 31). There was also a signifi-cant difference for hyperintense lesions (BH mean 6.076 SD 10.24, RT mean 4.13 6 SD 6.21, P ¼ 0.027, N ¼15). There was no significant difference in lesion detec-tion for isointense lesions. A score of 4 for lesionsharpness equated to a reference lesion that was notdetectable. For the RT sequence, there were 14 nonde-tectable lesions for reviewer 1 and 13 for reviewer 2compared with four lesions on BH for both reviewers.

Overall lesion to liver contrast ratio was significantlybetter for RT (mean 6 SD; for RT 0.26 6 0.24 com-pared with BH 0.21 6 0.20, P ¼ 0.046). When thelesions where stratified into intensities, no significantdifferences in lesion to liver contrast ratio was shownfor the hyperintense (0.13 6 0.17 for RT and 0.10 6

0.04 for BH, P ¼ 0.820) and isointense groups (mean 6

SD; for RT 0.03 6 0.02 compared with BH 0.02 6 0.01,P ¼ 0.286). RT had better contrast ratio for hypointenselesions however this was not statistically significant(0.39 6 0.23 for RT; 0.32 6 0.21 for BH; P ¼ 0.055).

DISCUSSION

With use of hepatocyte-specific contrast agents suchas gadoxetic acid, improved lesion characterizationand detection have been shown (1–3,9,10). This is dueto increased signal intensity of liver parenchyma in thehepatobiliary phase (11,12) contrasting with lack ofuptake in nonhepatocyte lesions. In addition, theuptake of contrast in hepatocyte containing lesionscan help with characterization (13). Due to the pro-longed enhancement of the liver parenchyma in the he-patocyte phase (14), sequences that are not as timecritical in acquisition have been developed in anattempt to further enhance the advantages of gadoxeticacid. Asbach reported on a novel respiratory triggeredinversion recovery prepared T1 weighted sequence (5)which produced better image quality then BH VIBE.The significant advantages of the RT sequence were

described as improved liver SNR, liver to lesion con-trast, increased sharpness of hepatic vessels and com-mon bile duct, better depiction of focal liver lesions,and sharpness of hepatic vessels and assessed biliarysegments. The level of artifact between RT and BH wasreported as not significantly different. In our study, wehave compared the RT sequence with the standardbreathhold VIBE GRE sequence in a larger populationgroup with wider inclusion criteria including lesions ofall intensities, not just hypointense lesions. The hy-pothesis being tested is that RT would improve SNR,lesion to liver contrast and allow high spatial resolutionimaging with the end result of improved lesion detec-tion. Disadvantages of the RT are increased acquisitiontime (5) as data are discarded except when the triggeris activated. Additionally, it may be a pulse sequencenovel to a particular vendor and not available univer-sally across vendors. On the other hand, a robust formof three-dimensional (3D) T1 GRE pulse sequence iscurrently universally available on all vendors.

Our study confirmed that the high resolution RTsequence is technically feasible. We noted a highermean SNR (not statistically significant) with the RTsequence and improved lesion to liver contrast (forhypointense lesions); however, this did not result inimproved lesion detection. Asbach reported a signifi-cant difference in SNR between the two sequences;however, we did not confirm this with our study. Thedifferences in SNR measurement methodology betweenthe two studies may account for the different results.Asbach used a ‘‘difference method’’ to calculate SNR inone volunteer patient without the administration ofcontrast. This study used the ‘‘SD method,’’ which canbe inaccurate for sequences using reconstruction fil-ters and parallel imaging when noise does not follow aRayleigh distribution (7). Our measurements, however,closely followed a Rayleigh distribution and hencecould be considered reasonably accurate.

We confirmed that RT had better lesion to liver con-trast when all intensities were considered. However,the difference of the mean contrast ratio is small andmay not be perceptible. This is supported by the

Figure 1. T1 weighted liver MRI in hepatobiliary phase using gadoxetic acid. Respiratory triggered sequence (a) demonstrat-ing artifact causing inhomogeneous liver parenchyma, partially obscuring hyperintense lesion in segment 2 (arrow) which isseen in breathhold, T1 weighted gradient echo sequence (b).

RT Versus BH GRE MRI Liver Using Gd-EOB-DTPA 703

qualitative rankings for lesion contrast which werenot significant for reviewer 1 and reviewer 2 rating BHbetter for lesion to liver contrast. The small differencein overall lesion to liver contrast is probably related toour wider inclusion criteria with hyper-, hypo-, andisointense lesions assessed. When only hypointenselesions were considered, consistent with Asbach’sstudy, RT had higher lesion to liver contrast valuesthan BH although this was not statistically signifi-cant. We hypothesize that liver parenchymal signal in-tensity is higher in RT sequences due to the increasedacquisition time and inversion recovery preparation.While this theoretically improves contrast for hypoin-tense lesions, this may adversely affect lesion contrastratio for hyperintense lesions.

For qualitative factors, the two reviewers rated BHsequence better for lesion sharpness, hepatic vesselsharpness, and level of artifact, which was in contrastto previously published results (5). Our study alsoassessed total number of lesions detected with signifi-cantly more hyper- and hypointense lesions detectedon BH sequence. We found that the RT sequenceimage quality was degraded due to increased artifact,particularly from respiratory motion and ghosting,

which has been reported in prior studies (15–18).When using respiratory triggering, these artifacts arenoted to be worse with irregular breathing patterns orhigh respiratory rates when there is more than one re-spiratory cycle per TR (19). Conversely, with regularbreathing patterns, decreased artifacts using respira-tory triggering have been noted in prior studies basedon T2 imaging (19,20). We found the artifacts affectedthe homogeneity of the liver parenchyma in the hepa-tobiliary phase with resultant decreased lesion detec-tion (Figs. 1 and 2). Artifacts also affected the dome ofthe liver (Fig 3), left lobe, and gallbladder fossa(Figs. 4 and 5), which could potentially decreasedetection of lesions.

Our study had several limitations. Two scanners at1.5T and 3.0T were used with the results then pooled.However, both sequences were performed only on onescanner for each patient. Different resolutions wereused between and within the RT and BH groups. Inall cases, the RT sequence had higher spatial resolu-tion than BH which would favor RT in lesion detec-tion. We did not consider the final pathological diag-nosis as a factor, as our aim was to compare the twosequences (BH and RT) on lesion detection and

Figure 2. Inhomogeneity of liver parenchyma in the respiratory triggered sequence (arrowheads in a) obscuring segment 3hyperintense lesion (arrows) which is seen in the breathhold sequence (b).

Figure 3. Low signal in the respiratory triggered sequence (a) affecting the dome of the liver and obscuring a segment 8hypointense lesion (black arrow) which is seen in breathhold sequence (b).

704 Tran and Jhaveri

intensity of lesions. Our results cannot be applieddirectly to a pathological group of lesions. Hypoin-tense lesions could be considered representative ofboth benign (e.g., cysts) and malignant lesions (e.g.,nonhepatocellular metastases). We had two reviewerswith potential for bias from reviewer 2 who wasinvolved in both quantitative and qualitative assess-ments. The chance of bias was reduced by anonymiz-ing patient and sequence data and randomizing cases.The qualitative review was also performed more than6 months after quantitative review. Only moderateagreement was achieved between the two reviewers inregards to qualitative factors. There is potential forinaccurate SNR measurement when multichannelcoils, reconstructive filters, and parallel imaging areused. Our signal intensity measurements closely fol-lowed a Rayleigh distribution so the ‘‘SD method’’ wasconsidered a valid estimation of SNR. Our limitedpopulation size and larger proportion of hypointenselesions may skew the outcomes.

High resolution T1 imaging with hepatocyte-specificcontrast agents has potential to improve detection of

liver lesions such as metastases (10,21,22) and HCC(23). With the use of novel sequences using respira-tory triggering, further improvement in image qualityis sort with the end result of better lesion detection.Earlier detection may influence patient managementand is some cases improve patient outcomes (24).Imaging also remains challenging in noncooperativepatients (6,25) and respiratory triggering may have arole providing an additional option for assessment. Inour study, we found that further development isrequired for a more robust sequence that reduces arti-fact compared with the standard breathholdsequence. Based on our results, we would advocatecontinued use of the BH VIBE sequence in patientswho are capable. In patients with limited respiratoryreserves, RT could be used after adequate instructionreinforcing a regular breathing cycle. In patients withhypointense metastases who are considered for cura-tive surgical or ablative therapies, high resolution RTsequence may also be advantageous.

In conclusion, our study showed that standard BHVIBE GRE T1 weighted sequence produced better

Figure 4. Artifact in the respiratory triggered sequence (a) affecting the left liver lobe and gallbladder fossa (arrows). Equiva-lent slice on the breathhold sequence (b) shows homogeneous parenchyma.

Figure 5. Artifact in the respiratory triggered sequence (a) affecting the gallbladder fossa (small arrows). Inhomogeneity ofthe left lobe (long arrowhead) partly obscures the low signal lesion with peripheral hyperintensity in segment 2 (small arrow-head) which demonstrates sharper margins on breathhold sequence (b). Heterogeneous fat saturation (asterisk).

RT Versus BH GRE MRI Liver Using Gd-EOB-DTPA 705

overall image quality than high resolution RT IR GRET1 weighted sequence in several quantitative andqualitative factors.

REFERENCES

1. Vogl TJ, Kummel S, Hammerstingl R, et al. Liver tumors: com-parison of MR imaging with Gd-EOB-DTPA and Gd-DTPA. Radiol-ogy 1996;200:59–67.

2. Reimer P, Rummeny EJ, Shamsi K, et al. Phase II clinical evalua-tion of Gd-EOB-DTPA: dose, safety aspects, and pulse sequence.Radiology 1996;199:177–183.

3. Huppertz A, Haraida S, Kraus A, et al. Enhancement of focal liverlesions at gadoxetic acid-enhanced MR imaging: correlation withhistopathologic findings and spiral CT–initial observations. Radi-ology 2005;234:468–478.

4. Yamashita Y, Yamamoto H, Namimoto T, Abe Y, Takahashi M.Phased array breath-hold versus non-breath-hold MR imaging offocal liver lesions: a prospective comparative study. J MagnReson Imaging 1997;7:292–297.

5. Asbach P, Warmuth C, Stemmer A, et al. High spatial resolutionT1-weighted MR imaging of liver and biliary tract during uptakephase of a hepatocyte-specific contrast medium. Invest Radiol2008;43:809–815.

6. Altun E, Semelka RC, Dale BM, Elias J Jr. Water excitationMPRAGE: an alternative sequence for postcontrast imaging of theabdomen in noncooperative patients at 1.5 Tesla and 3.0 TeslaMRI. J Magn Reson Imaging 2008;27:1146–1154.

7. Dietrich O, Raya JG, Reeder SB, Reiser MF, Schoenberg SO. Mea-surement of signal-to-noise ratios in MR images: influence ofmultichannel coils, parallel imaging, and reconstruction filters.J Magn Reson Imaging 2007;26:375–385.

8. Landis JR, Koch GG. The measurement of observer agreement forcategorical data. Biometrics 1977;33:159–174.

9. Ahn SS, Kim MJ, Lim JS, Hong HS, Chung YE, Choi JY. Addedvalue of gadoxetic acid-enhanced hepatobiliary phase MR imag-ing in the diagnosis of hepatocellular carcinoma. Radiology 2010;255:459–466.

10. Kim YK, Lee YH, Kwak HS, Kim CS, Han YM. Detection of livermetastases: Gadoxetic acid-enhanced three-dimensional MRimaging versus ferucarbotran-enhanced MR imaging. Eur JRadiol 2010;73:131–136.

11. Feuerlein S, Boll DT, Gupta RT, Ringe KI, Marin D, Merkle EM.Gadoxetate disodium-enhanced hepatic MRI: dose-dependentcontrast dynamics of hepatic parenchyma and portal vein. AJRAm J Roentgenol 2011;196:W18–W24.

12. Hamm B, Staks T, Muhler A, et al. Phase I clinical evaluation ofGd-EOB-DTPA as a hepatobiliary MR contrast agent: safety,pharmacokinetics, and MR imaging. Radiology 1995;195:785–792.

13. Goodwin MD, Dobson JE, Sirlin CB, Lim BG, Stella DL. Diagnos-tic challenges and pitfalls in MR imaging with hepatocyte-specificcontrast agents. Radiographics 2011;31:1547–1568.

14. Ringe KI, Husarik DB, Sirlin CB, Merkle EM. Gadoxetate diso-dium-enhanced MRI of the liver: part 1, protocol optimizationand lesion appearance in the noncirrhotic liver. AJR Am J Roent-genol 2010;195:13–28.

15. Lee HY, Lee JM, Kim SH, et al. Detection and characterization offocal hepatic lesions by T2-weighted imaging: comparison of navi-gator-triggered turbo spin-echo, breath-hold turbo spin-echo,and HASTE sequences. Clin Imaging 2009;33:281–288.

16. Klessen C, Asbach P, Kroencke TJ, et al. Magnetic resonanceimaging of the upper abdomen using a free-breathing T2-weighted turbo spin echo sequence with navigator triggered pro-spective acquisition correction. J Magn Reson Imaging 2005;21:576–582.

17. Lee SS, Byun JH, Hong HS, et al. Image quality and focal lesiondetection on T2-weighted MR imaging of the liver: comparison oftwo high-resolution free-breathing imaging techniques with twobreath-hold imaging techniques. J Magn Reson Imaging 2007;26:323–330.

18. Vasanawala SS, Iwadate Y, Church DG, Herfkens RJ, Brau AC.Navigated abdominal T1-W MRI permits free-breathing image ac-quisition with less motion artifact. Pediatr Radiol 2010;40:340–344.

19. Low RN, Alzate GD, Shimakawa A. Motion suppression in MRimaging of the liver: comparison of respiratory-triggered and non-triggered fast spin-echo sequences. AJR Am J Roentgenol 1997;168:225–231.

20. Pauleit D, Textor J, Bachmann R, et al. Improving the detectabil-ity of focal liver lesions on T2-weighted MR images: ultrafastbreath-hold or respiratory-triggered thin-section MRI? J MagnReson Imaging 2001;14:128–133.

21. Shimada K, Isoda H, Hirokawa Y, Arizono S, Shibata T, TogashiK. Comparison of gadolinium-EOB-DTPA-enhanced and diffu-sion-weighted liver MRI for detection of small hepatic metastases.Eur Radiol 2010;20:2690–2698.

22. Choi JY, Choi JS, Kim MJ, et al. Detection of hepatic hypovascu-lar metastases: 3D gradient echo MRI using a hepatobiliary con-trast agent. J Magn Reson Imaging 2010;31:571–578.

23. Sano K, Ichikawa T, Motosugi U, et al. Imaging study of early he-patocellular carcinoma: usefulness of gadoxetic acid-enhancedMR imaging. Radiology 2011;261:834–844.

24. Gillams AR, Lees WR. Five-year survival following radiofrequencyablation of small, solitary, hepatic colorectal metastases. J VascInterv Radiol 2008;19:712–717.

25. Azevedo RM, de Campos RO, Ramalho M, Heredia V, Dale BM,Semelka RC. Free-breathing 3D T1-weighted gradient-echosequence with radial data sampling in abdominal MRI: prelimi-nary observations. AJR Am J Roentgenol 2011;197:650–657.

706 Tran and Jhaveri