| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Technical Developments |
1 From the Departments of Radiology (J.A., O.V., P.L., H.G.) and Biostatistics (J.C.), Université René Descartes, Hôpital Cochin, 27 rue du Fg Saint Jacques, 75679 Paris Cedex 14, France; and GE Medical Systems, Buc, France (C.A.). Received June 7, 2001; revision requested July 3; revision received October 18; accepted December 10. Address correspondence to O.V. (e-mail: olivier.vignaux@cch.ap-hop-paris.fr).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
© RSNA, 2002
Index terms: Liver neoplasms, MR, 761.121411 • Magnetic resonance (MR), comparative studies, 761.121411 • Magnetic resonance (MR), pulse sequences, 761.121411
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
Thus, the challenge is to develop a breath-hold sequence that can be performed rapidly, maintain the high level of contrast provided by conventional triggered fast SE sequences, reduce image artifacts, and be recommended as a routine substitute for conventional fast SE imaging. A modified fast SE technique with fast recovery was developed (6) to allow full coverage of the liver in one or two breath holds with high-quality T2-weighted images. Initial experience in the liver was promising and resulted in image quality and hepatic lesion detection values that were higher than those with the standard fast SE technique (6). The aim of this study was to compare this breath-hold fast-recovery fast SE sequence with breath-hold half-Fourier single-shot fast SE and non–breath-hold respiratory-triggered fast SE sequences on the basis of image quality and hepatic lesion conspicuity on T2-weighted MR images.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
The patient group comprised 23 men and 15 women (age range, 28–78 years; mean age, 53 years). Twenty-seven patients had focal abnormalities that corresponded to a total of 123 hepatic lesions (diameter range, 0.8–9.0 cm) (number of focal abnormalities per patient: one, n = 15; two to five, n = 7; five to 10, n = 3; more than 10, n = 2). Lesion diagnoses included metastasis (n = 36), hepatocellular carcinoma (n = 23), hemangioma (n = 31), cysts (n = 27), focal nodular hyperplasia (n = 5), and liver purpura (n = 1). Proof of malignant lesions had been obtained by means of surgery (resection or surgical biopsy), percutaneous liver biopsy, or the observance (in the diagnosis of hepatocellular carcinoma) of a marked elevation in 
-fetoprotein levels after follow-up CT depicted growth of the hepatic tumor. Diagnoses of focal nodular hyperplasia, cavernous hemangioma, or cysts were made on the basis of the patterns seen at dynamic gadolinium-enhanced MR imaging compared with the typical appearance seen at CT and duplex Doppler US. The diagnosis of purpura was established by means of pathologic tissue examinations. In 11 cases, no focal abnormalities of the liver parenchyma were found on MR images.
Breath-hold Fast-Recovery Fast SE Sequence
Modifications to the fast SE pulse sequence included fast-recovery T2 enhancement and an optimized section-ordering scheme. Fast-recovery T2 enhancement was added to this sequence to increase the signal from long T2 components in the liver. The breath-hold fast-recovery technique makes use of additional radio-frequency pulses after the final acquisition window to drive the recovery of longitudinal magnetization (Fig 1). An additional 180° refocusing pulse is played out once the last echo in the fast SE echo train has been acquired. A -90° pulse is then used to drive the refocused magnetization back up onto the longitudinal axis instead of allowing it to recover with the T1 processes. After an interval of several repetition times, a steady state of longitudinal magnetization is established with net enhancement of the long T2 components. For long echo trains, signals from fluid can be enhanced dramatically in these images.
|
With use of the breath-hold fast-recovery method, transverse sections were acquired in groups of contiguous sections by starting with the first group located at the top of the liver; each subsequent group was gradually located lower, until the entire liver was covered. (Groups of contiguous sections can also be acquired by starting from the bottom of the liver and progressing to the top during subsequent breath holds, if preferred.) One group of sections was acquired for each breath hold, and the sections within each group were interleaved during the breath hold to reduce cross-talk effects. The ability to overlap groups of sections may prevent the possibility of missing tissue between groups of sections if the patient does not hold his or her breath at the same location during subsequent breath holds.
MR Imaging Protocol
MR imaging was performed with a superconducting 1.5-T imager (Signa Lx; GE Medical Systems, Milwaukee, Wis) with a phased-array coil as the receiver. All images were obtained in the transverse plane. Because the results of previous studies (1,5,7,8) demonstrated the benefits of using the fat-suppression technique to help detect liver tumors, all acquisitions were fat suppressed. Fat suppression was achieved (FatSat; GE Medical Systems) with a spectrally selective radio-frequency pulse. The section thickness was 9 mm with all pulse sequences. To provide images of optimum quality and full liver coverage in one acquisition with the respiratory-triggered fast SE sequence and in one breath hold with the single-shot fast SE sequence, a 2-mm intersection gap was applied. After application of the ordering scheme included in breath-hold fast-recovery imaging, sections were acquired in two breath holds and were contiguous.
The T2-weighted imaging protocol included (a) a conventional respiratory-triggered fast SE sequence (repetition time msec [depending on respiratory intervals]/echo time msec of 3,000–4,000/90, 256 x 256 matrix, 31.25-kHz received bandwidth, 38 x 28-cm field of view, echo train length of 16, 30% respiratory trigger point, 40% trigger window), (b) the breath-hold fast-recovery fast SE sequence (3,000/90, 256 x 256 matrix, 31.2-kHz received bandwidth, 38 x 28-cm field of view, echo train length of 19), and (c) a breath-hold half-Fourier single-shot fast SE sequence (>30,000/90, 256 x 256 matrix, 31.25-kHz received bandwidth, 38 x 28-cm field of view).
We subsequently performed T1-weighted fast SE imaging (600/10 with 256 x 192 matrix) and triphasic dynamic gadolinium-enhanced (Magnevist; Berlex Laboratories, Wayne, NJ) gradient-echo imaging (225/minimal with 80° flip angle) in all patients. In all MR imaging examinations, spatial presaturation pulses were used above and below the imaging volume.
Quantitative Image Analysis
Quantitative analysis was performed on T2-weighted MR images obtained with the three pulse sequences by using operator-defined (J.A.) region-of-interest measurements of mean signal intensity in the liver, background noise, and hepatic lesions, when present. Signal intensity in the liver was measured in areas devoid of focal changes in signal intensity, as well as in large vessels or prominent artifacts. For liver lesions, a circular region of interest was drawn to encompass as much of the lesion as possible. In patients with multiple lesions, a single lesion representative of the average size of the patient’s lesions was chosen on the contrast material–enhanced (Omnipaque; Nycomed Amersham, Princeton, NJ) CT scan obtained within 10 days before the MR examination. When a lesion was suspected at preliminary US but was not visible on the CT scan, the most conspicuous lesion was chosen from the MR examination, and signal intensity measurements were obtained for this lesion on images obtained with each of the T2-weighted pulse sequences. When the hepatic lesion was too small to place a region of interest, the image was magnified as much as three times. However, only lesions with a diameter of 10 mm or more were included for quantitative evaluation to exclude any inaccuracies in signal intensity measurements that resulted from partial volume effects in smaller lesions.
Whenever possible, regions of interest that were at least 35 mm2 were used. The SD of background noise was measured in the largest possible region of interest positioned in the phase-encoding direction outside the abdominal wall, to account for any motion artifacts. Then, the following data were calculated: liver signal-to-noise ratio (SNR), (SIliver/SDnoise); and lesion-to-liver contrast-to-noise ratio (CNR), ([SIlesion - SIliver]/SDnoise); where SI is signal intensity.
Qualitative Image Analysis
Image quality with each of the three T2-weighted sequences was assessed separately by two independent readers (J.A., O.V.), who interpret MR images of the liver as part of their daily clinical and research practice. Both radiologists were blinded to the presence or absence of disease. After they completed their individual reviews, the observers then reviewed the images jointly and reached a consensus for all qualitative parameters. Overall image quality and visualization of intrahepatic vessels (including patient motion, respiratory ghosting, vascular pulsation, peristalsis, and susceptibility artifacts) were graded on a five-point scale: 1, unacceptable; 2, poor; 3, fair; 4, good; 5, excellent. Image artifacts were also graded on a five-point scale: 1, unacceptable; 2, severe; 3, moderate; 4, mild; 5, absent. Whenever focal hepatic lesions were present, lesion detectability and the ability to determine the size of the lesion were also evaluated. For each technique, the observers recorded the site (Couinaud segment) of visible abnormalities and indicated the presence of solid or nonsolid lesions.
All MR images were then reviewed jointly by two other radiologists (H.G., P.L.), who served as study coordinators. In the event of a discrepancy between the three MR pulse sequences (ie, when a focal lesion was visible on images obtained with only one or two sequences), preliminary triphasic helical CT and US scans were reviewed to obtain a consensus. Lesion detection was then graded on the basis of the sequence that best demonstrated the liver lesions. One hundred twenty-three hepatic lesions were analyzed. Fifty-nine were nonsolid lesions (including cysts, hemangiomas, and liver purpura), and 64 were solid lesions (including hepatocellular carcinoma, liver metastases, and focal nodular hyperplasia). Evaluations of the ability to determine the size of the lesion were based on the sharpness of lesion margins, which enabled accuracy of measurement of lesion size.
Statistical Analysis
Data were reported as the mean ± 1 SD. Owing to a nonnormal distribution of most variables and to the small number of subjects in groups of interest, nonparametric statistical methods (pairwise Wilcoxon test) were used to compare the breath-hold fast-recovery fast SE sequence with conventional and single-shot fast SE sequences. Two-tailed P values were calculated. Differences with P < .05 were considered significant.
The concordance or agreement of observers for qualitative analysis was further assessed by calculating 
statistics to validate the scoring method.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
|
= 0.82).
|
Considering all lesions detected, 87 were located in the right side of the liver (including 18 lesions in segment VIII, 22 in segment VII, 18 in segment V, and 29 in segment VI). Thirty-six lesions were located in the left side of the liver (including 17 lesions in segment IV, 10 in segment III, and nine in segment II).
Although all fast SE studies were deemed to provide diagnostically adequate lesion detectability, the overall quality rating was inferior to that with the single-shot fast SE sequence (P = .005) (Fig 2). This difference was statistically significant for solid lesions (P < .05) (Fig 3) but not for cystic lesions. The breath-hold fast-recovery and respiratory-triggered fast SE pulse sequences were comparable in terms of detectability of both solid and nonsolid lesions.
|
|
|
|
|
|
The breath-hold fast-recovery sequence provided fewer image artifacts than did the respiratory-triggered fast SE sequence (Fig 4); the difference was statistically significant (P = .002). In terms of the two breath-hold techniques, the reviewers’ scores were slightly higher for the single-shot fast SE sequence, but the difference was not significant. The depiction of intrahepatic vessels did not differ significantly among the three techniques. The overall image quality of breath-hold fast-recovery images was significantly superior to that of conventional (P = .002) and single-shot fast SE (P = .01) images.
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
By refocusing residual magnetization in the transverse plane and flipping it back to the longitudinal axis, the modified fast SE sequence with fast recovery provides T2 weighting and signal that are improved compared with those with a conventional fast SE sequence. Fast recovery is added to this pulse sequence not only to increase T2 contrast on the images but also to enable either shorter breath-hold acquisitions, by allowing a reduction in repetition time, or a larger matrix, by using reasonable breath-hold timing.
SNRs for the liver were higher with breath-hold fast-recovery than with conventional fast SE sequences, probably because the noise related to motion was extremely low. Although the respiratory motion that occurs during a T2-weighted non–breath-hold imaging sequence is known to cause loss of signal from lesions, we found higher lesion signals with respiratory-triggered T2-weighted than with breath-hold fast-recovery sequences. A longer acquisition time may account for this difference. However, the higher noise levels on respiratory-triggered fast SE images contribute to the lower CNRs calculated for non–breath-hold images.
With respect to water-rich lesion detection, no significant differences were observed among the three sequences, although the breath-hold fast-recovery technique depicted three more lesions than did the respiratory-triggered fast SE technique, probably because of reduced spatial resolution related to ghosting artifacts on the conventional fast SE images. In two cases, small cystic lesions were missed on breath-hold single-shot fast SE images, probably because the lesion was located in an intersection gap. Contiguous single-shot fast SE acquisitions during two breath holds do not prevent the possibility of missing tissue between groups of sections if the patient does not hold his or her breath at the same location during the subsequent breath holds.
In contrast, the optimized ordering scheme of the breath-hold fast-recovery technique allowed complete liver coverage in all cases in two breath holds, with contiguous sections acquired in an interleaved fashion to prevent the misdiagnosis of small lesions in an intersection gap. It has been suggested that respiratory motion that occurs during non–breath-hold fast SE acquisition may be useful in these cases, because small lesions that could be located in an intersection gap at the time of the image prescription may be present in imaged sections for some portion of the acquisition (10).
In the case of solid lesion detection, we found inferior diagnostic capability with single-shot fast SE images. These findings are compatible with those of Hori et al (11) and have been related to image blurring. One of the different sources of such blurring in single-shot approaches is a loss of signal for the larger k-space values usually acquired with late echoes, with T2-filtering effects (12–14) and the "loss of small objects" (13–15). This phenomenon may therefore account for solid lesions that remain undetected because of their T2, which is shorter than that of cystic focal changes. Furthermore, it may reduce accuracy of the characterization of certain hepatic lesions.
A further explanation for the lower accuracy of solid lesion detection with single-shot fast SE imaging is the T1 contrast inherent in fast SE techniques, which is emphasized by half-Fourier single-shot fast SE sequences, because of the stimulated echo contribution to the signal. This phenomenon is attenuated in breath-hold fast-recovery imaging because additional pulses occur at the end of the echo train that refocus the residual magnetization and prevent it from recovering with a T1 process.
On the basis of our results concerning image quality, breath-hold techniques were best for presenting fewer image artifacts. We found that the appearance with two breath-hold sequences (breath-hold fast recovery and single-shot fast SE) differed considerably. Single-shot fast SE images were free of artifacts, but the overall image quality was impaired by blurring. In contrast, breath-hold fast-recovery images revealed better anatomic and lesion sharpness, which resulted in improved ability to determine the size of the lesion. In the case of respiratory-triggered sequences, images were of high quality in terms of lesion contours, but phase-shift artifacts affected the overall image quality.
There were certain limitations to this study. First, in patients with focal liver abnormalities, the true number of lesions was unknown because pathologic results were not available in most cases. However, this bias probably did not dramatically influence lesion detectability, since this parameter was judged on the basis of comparison among different pulse sequences. Furthermore, this study was not focused on characterization of lesions that would enable a reduction in the need for systematic pathologic proof.
Another potential bias was related to the subjectivity of qualitative evaluation, which was emphasized by the fact that the reviewers were not really blinded to which pulse sequence was evaluated because obvious differences existed between the images obtained with the three techniques.
In summary, the results of this study suggest that in a short acquisition time, the breath-hold fast-recovery technique provides more satisfactory contrast for both solid and cystic masses with better anatomic sharpness and higher soft-tissue resolution than does the breath-hold single-shot fast SE technique. In addition, the breath-hold fast-recovery technique provides a dramatically lower level of image artifacts than does a conventional respiratory-triggered pulse sequence. Despite the limited number of patients included in this initial evaluation, our results are encouraging, and breath-hold, optimized, fast-recovery, fast SE MR imaging appears to constitute an attractive alternative to both conventional respiratory-triggered and breath-hold single-shot fast SE sequences for first-line T2-weighted MR imaging of the liver.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, J.A., O.V., P.L.; study concepts and design, J.A., O.V., P.L.; literature research, J.A.; clinical studies, J.A., O.V., H.G., P.L.; data acquisition, J.A., O.V., H.G., P.L.; data analysis/interpretation, J.A., J.C.; statistical analysis, J.C.; manuscript preparation, J.A., O.V.; manuscript definition of intellectual content, J.A., O.V., P.L.; manuscript editing, J.A., O.V., H.G.; manuscript revision/review, O.V., P.L., J.C., C.A.; manuscript final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. D. Chang and R. F. Thoeni Effect of T1 Shortening on T2-Weighted MRI Sequences: Comparison of Hepatic Mass Conspicuity on Images Acquired Before and After Gadolinium Enhancement Am. J. Roentgenol., May 1, 2008; 190(5): 1318 - 1323. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Parikh, S. J. Drew, V. S. Lee, S. Wong, E. M. Hecht, J. S. Babb, and B. Taouli Focal Liver Lesion Detection and Characterization with Diffusion-weighted MR Imaging: Comparison with Standard Breath-hold T2-weighted Imaging Radiology, March 1, 2008; 246(3): 812 - 822. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Huang, S. S. Raman, N. Vuong, J. W. Sayre, and D. S. K. Lu Utility of Breath-Hold Fast-Recovery Fast Spin-Echo T2 Versus Respiratory-Triggered Fast Spin-Echo T2 in Clinical Hepatic Imaging Am. J. Roentgenol., March 1, 2005; 184(3): 842 - 846. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
/90) transverse fast SE T2-weighted MR images in a patient with a small peripheral liver hemangioma (arrow). Depiction and sharpness of the lesion are best in b. In c, the lesion is partially located in an intersection gap, which results in a partial volume effect and thus a loss of signal intensity and lesion contours. A blurring effect may also account for this aspect. Both reviewers mistook the lesion in c for an intrahepatic anatomic structure.
