| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full paper |
1 Department of Radiology, AZ St-Jan AV, Bruges, Belgium, 2 Philips Medical Systems, Best, The Netherlands, 3 Hospital Administration and Statistics, AZ St-Jan AV, Bruges, Belgium, 4 Department of Radiology, Academic Medical Centre, Amsterdam, The Netherlands
Correspondence: Kenneth CoenegrachtsMD, Department of Radiology, AZ St-Jan AV, Ruddershove 10, B-8000 Bruges, Belgium. E-mail: kenneth.coenegrachts{at}azbrugge.be
 |
Abstract |
|---|
 Top Abstract IntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Recently, diffusion-weighted imaging techniques for the identification of focal liver lesions have been examined [6, 7]. Using diffusion-weighted single-shot spin echo echoplanar imaging (SS SE-EPI DWI), black blood images of the liver were obtained by applying low b-values, facilitating differentiation between a lesion and a vessel [6, 8].
The purpose of this study was to compare, qualitatively and quantitatively, respiratory-triggered SS SE-EPI DWI sequences with four b-values (b = 0, b = 20, b = 300, b = 800 s mm–2) and T2W SS TSE in patients with focal liver lesions, with special focus on small (<10 mm) lesions. The potential of SS SE-EPI DWI in the identification of different types of focal liver lesions (biliary cysts, haemangiomas and metastases) with different sizes (<10 mm, 10–20 mm and >20 mm) was compared with T2W SS TSE.
|
Methods and materials |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
This prospective study was approved by the hospital ethics committee, and written informed consent was obtained from all patients.
Technique
A 1.5 T MRI whole-body scanner (Intera, Philips Medical Systems, Best, The Netherlands) with a four-element SENSE (SENSitivity Encoding) body phased-array coil was used. The respiratory-triggered T2W SS TSE and respiratory-triggered SS SE-EPI DWI images were acquired using the following parameters:
Analysis
The different types of detected focal liver lesions (biliary cysts, haemangiomas and metastases) are subgrouped by size (<10 mm, 10–20 mm and >20 mm) for further analysis.
All examinations were read on a picture archiving and communication system (PACS) workstation (Agfa, Mortsel, Belgium) allowing the readers to scroll up and down the data set to evaluate whether a suspected lesion could be mistaken for a vascular structure.
Each lesion was detected without bias and, once detected, the location was recorded by writing down the number of the slice in which a detected lesion was found. Matching was possible by comparing the slice numbers of each sequence of the different readers.
Qualitative analysis
Two abdominal radiologists, experienced in interpreting liver MRI in daily clinical practice (5 years' and 13 years' experience, respectively), independently evaluated all images and subjectively rated each sequence for overall image quality, lesion conspicuity and artefacts. To avoid any learning bias, review of each image was done in a randomized, blinded fashion. Overall image quality and lesion conspicuity were based on the following five grading scales: excellent = 1; good = 2; fair = 3; poor = 4; and unacceptable = 5. The presence of artefacts was rated using the following four grading scales: absent = 1; mild = 2; moderate = 3; and severe = 4. The two radiologists did not have any other information about patient history, clinical examination, laboratory results, findings of other imaging techniques or final diagnosis.
Quantitative analysis
Lesion identification
Discrepancies of interpretation regarding the presence or absence of a lesion were resolved by means of a consensus reading by the same abdominal radiologists. The number of lesions was recorded in each sequence. For lesion identification in each image, the lesions on each image were evaluated by comparing them with the reference standard findings including ultrasound findings, follow up CT and contrast-enhanced MRI (see paragraph on Reference standard). For comparison of lesion identification between T2W SS TSE and SS SE-EPI DWI, only the maximum number of discrete lesions identified on one sequence was used to determine the total number of discrete lesions. Lesions not identified with any of the four remaining sequences were rated with the worst score for evaluation of the overall image quality and lesion conspicuity (score "5") and for evaluation of image artefacts (score "4") (see paragraph on Qualitative analysis).
Region of interest placement
The regions of interest (ROIs) were placed independently by both readers on the hepatic parenchyma and on all the focal liver lesions. An interval of at least 4 weeks was maintained between consensus reading for lesion identification and ROI placement. The ROIs on liver parenchyma were placed by avoiding intrahepatic vessels and intrahepatic lesions and consisted of at least 100 pixels. These ROIs were always placed in the posterior sector of the right liver lobe to avoid artefacts from the great vessels. For the liver lesions, a ROI was drawn manually to encompass the whole lesion. For heterogeneous lesions, ROIs included entire lesions, without separating components with various signal intensities.
Signal intensity measurements
Signal intensities were measured blinded to the measurements and evaluations performed by the other reader. To minimize any learning bias, we set the intervals for reviewing the five types of imaging protocols, that is respiratory-triggered T2W SS TSE and SS SE-EPI DWI with b = 0, b = 20, b = 300 and b = 800 s mm–2, at 2 weeks. Lesion-to-liver contrast-to-noise ratio (CNR) was calculated using the following equation:
|
|
Lesions that were not visible on a given pulse sequence (using only the studied T2W SS TSE sequence and the SS SE-EPI DWI sequence with b-values of 0, 20, 300 and 800 s mm–2) were rated with CNR = 0.
Reference standard
For biliary cysts and haemangiomas, the diagnosis was based on typical findings on ultrasound, CT or MRI. Typical lesion characteristics had to be present on at least two of the three imaging modalities.
For evaluating liver metastases in patients eligible for surgery, intraoperative ultrasound findings during surgery and histopathological findings were used as a reference standard. In all other patients, the final diagnosis was established by independent reading of all available imaging examinations (retrospective and prospective analysis) from all available imaging studies (ultrasound, CT and MRI) by two radiologists, and follow-up imaging was used. MRI included T2W SS TSE, T2W SS TSE with fat saturation, inversion recovery T1W gradient echo, in and out phase imaging and fat-saturated T1 weighted three-dimensional (T1W 3D) gradient echo imaging before the injection of gadolinium (Gd)-BOPTA , during the arterial, portal–venous and late venous phase and during the delayed phase (1–1.5 h after the injection of Gd-BOPTA). Findings at SS SE-EPI DWI were not taken into account. No consensus reading was needed for this final evaluation as no differences existed in the evaluation of the images between readers. To differentiate between a focal liver lesion and an artefact, all patients had a follow-up CT or MRI examination at least 6 months after the SS SE-EPI DWI.
Statistical analysis
RIDIT analysis was originally developed by Bross for the analysis of ordinal data [9]. RIDIT analysis calculates one aggregate score that is the probability of a higher/lower score in the distribution under investigation (e.g. image quality of T2W images) relative to a common reference distribution (e.g. image quality of all images irrespective of the technique) [9–12]. The null hypothesis is an a priori RIDIT of 0.5, which implies a 50/50 distribution. The RIDITs were all subtracted by 0.5 to have the mean at zero and multiplied by (–1) to have positive values for the promising results. The more positive the RIDIT, the better the result; the more negative, the worse the result for the considered imaging protocol. A RIDIT of zero means that the distribution of values of the criterion under consideration in the subgroup (e.g. T2W) is not different from the distribution of values in the reference population (all techniques). A difference was considered to be statistically significant at p<0.05.
For RIDIT analysis, Microsoft Excel 2000 (version 5.0, Washington, USA ) was used and for Kruskal–Wallis test, SPSS 12.0. (Windows, SPSS, Chicago, IL).
RIDIT analysis is not a readily available function in Excel, but the authors programmed it themselves and it is available from them on request.
As is well known, the "best" gold standard for lesion detection is intraoperative ultrasound with resection and histopathological analysis. Because of inoperable disease or because of the detection of merely benign liver lesions, this gold standard could not be performed in many cases. So the authors could not study the sensitivity or the specificity. As a consequence, receiver operating characteristic (ROC) analysis could not be performed.
|
Results |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Biliary cysts and haemangiomas
53 hepatic lesions in 8 patients were haemangiomas, and 36 lesions in 18 patients were biliary cysts. The diameter of haemangiomas and biliary cysts measured on the MR images ranged from 4 to 72 mm (mean diameter ± SD: 14.5 mm ± 10.3 mm) and from 3 to 29 mm (mean diameter ± SD: 15.0 mm ± 9.0 mm), respectively.
Liver metastases
40 lesions in 14 patients were colorectal liver metastases; five of these liver metastases were diagnosed by intraoperative ultrasound findings during surgery and histopathology findings. Diagnosis of the remaining 35 was determined based on all available imaging examinations and follow-up imaging after at least 6 months. The diameter of the liver metastases ranged from 4 to 26 mm (mean diameter ± SD: 10.9 mm ± 6.2 mm).
Qualitative analysis
Image quality
Figure 1a
shows significantly (p = 2.7 x 10–11) better overall image quality for T2W SS TSE compared with all SS SE-EPI DWI sequences. SS SE-EPI DWI with b = 20 s mm2 (p = 2.4 x 10–8) was significantly the best SS SE-EPI DWI sequence.
|
). A significantly (p = 4.5x10–12) better lesion conspicuity was obtained with T2W SS TSE for biliary cysts and with SS SE-EPI DWI with b = 20 s mm–2 for haemangiomas (p = 1.3x10–85) and metastases (p = 3.0x10–102) (Figure 1c). A significantly (p = 3.2x10–16) better lesion conspicuity was obtained with T2W SS TSE for biliary cysts <10 mm and with SS SE-EPI DWI with b = 20 s mm–2 for haemangiomas (p = 1.7x10–92) and metastases <10 mm (p = 2.4x10–105) (Figure 1d).
The usefulness of the coinciding black blood effect for the identification of focal liver lesions in the vicinity of the intrahepatic vasculature is shown in Figure 2.
|
).
|
shows the number of lesions identified on T2W SS TSE and SS SE-EPI DWI with b-values of 0, 20, 300 and 800 s mm–2. A substantial number of the focal liver lesions included were <10 mm in diameter. This is important for an accurate comparison of the studied sequences for the purpose of lesion identification. Four additional small (<10 mm) liver metastases were identified by SS SE-EPI DWI with b = 20 s mm–2 compared with (the second best imaging sequence) T2W SS TSE. Two of those additionally detected small liver metastases (<10 mm) were resected during surgery with histopathological proof; the other two additionally detected small liver metastases showed an increase in diameter during follow-up studies and were also detected on the additionally performed MRI sequences after intravenous injection of Gd-BOPTA. The increased detection of small lesions with the SS SE-EPI DWI sequence was thus not accompanied by an increase in false positives.
|
Quantitative analysis
The mean lesion-to-liver CNR among biliary cysts, haemangiomas and metastases in liver for T2W SS TSE and SS SE-EPI DWI images with b = 0 s mm–2 and b = 20 s mm–2 are summarized in Table 2. Comparison with SS SE-EPI DWI with b = 300 s mm–2 and b = 800 s mm–2 was difficult because of hampered visualization of several liver lesions due to suboptimal signal-to-noise and more pronounced artefacts.
|
For all metastases, the mean lesion-to-liver CNR was highest on SS SE-EPI DWI with a b-value of 20 s mm–2 compared with T2W SS TSE and SS SE-EPI DWI images with b = 0 s mm–2 (z = 11.17 and z = 11.82, respectively).
For biliary cysts <10 mm, significant differences in lesion-to-liver CNR were seen between SS SE-EPI DWI with a b-value of 20 s mm–2 and T2W SS TSE (z = 6.28) and between SS SE-EPI DWI with a b-value of 0 s mm–2 and 20 s mm–2 (z = 6.56). In this evaluation, T2W SS TSE showed the highest lesion-to-liver CNR.
For haemangiomas <10 mm, significant differences in lesion-to-liver CNR were seen between SS SE-EPI DWI with a b-value of 20 s mm–2 and T2W SS TSE (z = 8.44) and between SS SE-EPI DWI with a b-value of 0 s mm–2 and 20 s mm–2 (z = 8.68). In this evaluation, SS SE-EPI DWI (b = 20 s mm–2) showed the highest lesion-to-liver CNR.
For metastases <10 mm, significant differences in lesion-to-liver CNR were seen between SS SE-EPI DWI with a b-value of 20 s mm–2 and T2W SS TSE (z = 8.08) and between SS SE-EPI DWI with a b-value of 0 s mm–2 and 20 s mm–2 (z = 8.81). In this evaluation, SS SE-EPI DWI (b = 20 s mm–2) showed the highest lesion-to-liver CNR.
|
Discussion |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
For diffusion-weighted imaging, SS SE-EPI DWI with b = 20 s mm–2 shows the best results for the identification of focal liver lesions and lesion-to-liver CNR, especially for small (<10 mm) focal liver lesions. Overall, SS SE-EPI DWI with b = 300 s mm–2 and even more with b = 800 s mm–2 is not accurate in this study for the identification of focal liver lesions.
The presence of artefacts is explained by a severe chemical shift due to a rather low bandwidth in the phase-encoding direction of EPI sequences, and infolding which can be found on locations determined by the SENSE factor [13] and RFOV. Susceptibility artefacts of the lungs and surrounding air-filled bowel loops hampered the evaluation in those lesions that were located at the periphery of the liver and in the subphrenic hepatic areas [14].
The application of diffusion-weighted EPI for liver lesion identification and characterization is still not widely used. Mostly, emphasis has been on the calculation of parameters such as true diffusion, perfusion factor and apparent diffusion coefficient [15–19]. However, new interest has arisen in the potential of diffusion-weighted EPI for the identification of focal liver lesions [6, 8]. To our knowledge, this is the first study using diffusion-weighted imaging for the identification of small (<10 mm) liver lesions. The use of the black blood effect for facilitating identification of liver lesions has been described previously [8]. This coincident black blood effect mainly when using a b-value of 20 s mm–2 was useful in this study for identifying lesions <10 mm in diameter. The identification of small (<10 mm) focal liver lesions was the main purpose of this study.
In a review article by Morana et al [4], the clinical relevance of small lesion detection is discussed. Morana et al state that, when performing non-invasive imaging techniques such as ultrasound, CT and MRI (even when using intravenous contrast agents), focal liver lesions are mostly only detected down to a diameter of 10 mm. However, they showed when performing cadaver studies, when dealing with a primary colorectal carcinoma, that on average at least one liver metastasis <10 mm in diameter is missed for each detected liver metastasis with a diameter above 10 mm. The detection of even the smallest liver lesions is thus clinically important.
In the presented SS SE-EPI DWI sequence, the main importance of parallel imaging was the shortening of the echo train, thus reducing blurring effects [20].
Absence of the confusing bright signals from the intrahepatic vessels in this study facilitated the identification of focal liver lesions, particularly of small (<10 mm) liver metastases situated against intrahepatic vessels. No false-positive results were observed in this study.
As the "best" gold standard for lesion detection – being intraoperative ultrasound with resection and histopathological correlation – could not be performed in many cases, reading of all available imaging examinations (retrospective and prospective analysis of all available imaging studies (ultrasound, CT, MRI)) and follow-up imaging were used as a reference. Regarding the above explanation, sensitivity, specificity and accuracy cannot be calculated.
The RIDIT analysis is a powerful and elegant method for quantifying the different imaging techniques based on the visual scoring of the images, as is commonly done in clinical practice by radiologists.
Table 1 gives the lesion count to illustrate the number of detected lesions. As a detection means a CNR, it is our variable of interest. Consequently, Kruskal–Wallis tests are calculated on these CNR values and reveal a statistically significant difference among the techniques. CNR was measured using different techniques (T2W, b = 0, b = 20, b = 300, b = 800). We consider these techniques as independent from each other. For these reasons, the Kruskal–Wallis test was used. Comparing contrasts with the Kruskal–Wallis test produces z-scores that indicate significance.
A potential limitation could be the use of the presented T2W TSE sequence as a comparison for the diffusion-weighted SS SE-EPI sequence. The authors preferred to use a respiratory-triggered SS TSE rather than a breath-holding TSE to avoid cross-talk artefacts, to increase the signal-to-noise ratio and for patient comport. The NSA was 1 in the acquired T2W SS TSE sequence. Increasing the NSA might increase the intrinsic contrast of a TSE sequence, but also increases the risk of irregular breathing and movement of the patient. These movements may decrease image quality and increase the false-positive detection of focal liver lesions. The comparison of the presented diffusion-weighted SS SE-EPI sequence with a dedicated MRI sequence in combination with the use of superparamagnetic iron oxide (SPIO) could be a useful study for evaluating further the performance of SS SE-EPI DWI in the identification of focal liver lesions.
In this study, SS SE-EPI DWI was not examined for the accuracy of characterizing the different focal liver lesions. Further study needs to be done regarding the potential of SS SE-EPI DWI to characterize different focal liver lesions.
After consensus reading, ROI placements were performed by the same two radiologists. This could have had an influence on the ROI placement on the identified focal liver lesions. However, the drawing of the different ROIs was performed independently by two radiologists with a time interval of 4 weeks, resulting in independent signal intensity measurements.
The retrospective and prospective analysis of all available imaging studies was performed by the same two abdominal radiologists for establishment of the reference standard for the identification of focal liver lesions. This could have had a bias on the identification of the focal liver lesions. However, the purpose of this evaluation was to optimize the reference standard for lesion identification. Therefore, both retrospective and prospective analyses of all available images were preferred.
|
Conclusion |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
Received for publication May 17, 2006. Revision received October 4, 2006. Accepted for publication October 22, 2006.
|
References |
|---|
|
TopAbstractIntroductionMethods and materialsResultsDiscussionConclusionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  W. Luboldt, R. Kufer, N. Blumstein, T. L. Toussaint, A. Kluge, M. D. Seemann, and H.-J. Luboldt Prostate Carcinoma: Diffusion-weighted Imaging as Potential Alternative to Conventional MR and 11C-Choline PET/CT for Detection of Bone Metastases Radiology, December 1, 2008; 249(3): 1017 - 1025. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Bruegel, J. Gaa, S. Waldt, K. Woertler, K. Holzapfel, B. Kiefer, and E. J. Rummeny Diagnosis of Hepatic Metastasis: Comparison of Respiration-Triggered Diffusion-Weighted Echo-Planar MRI and Five T2-Weighted Turbo Spin-Echo Sequences Am. J. Roentgenol., November 1, 2008; 191(5): 1421 - 1429. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
