Magnetic Resonance Cholangiopancreatography in 3 Tesla : 2 D MRCP vs 3 D MRCP in Diagnostic Evaluation with Special Reference to Different Acquisition and Reconstruction Planes

MAGNETIC resonance cholangiopancreatography (MRCP) is an established technique for the evaluation of intraand extra-hepatic bile ducts in patients with known or suspected hepatobiliary disease [1]. It is considered a reliable, non-invasive alternative to diagnostic endoscopic retrograde cholangiopancreatography (ERCP) [2,3]. Since the first description by Wallner and colleagues in 1991 [4], different acquisition techniques have evolved. Most current MRCP techniques are based on heavily T2-weighted fast spin echo (FSE) pulse sequences, which yield a luminal image of the bile ducts that is based on the inherent signal of slowflowing or stationary bile. Both single-shot projections and multislice techniques are available [5], with the latter being distinguished into 2D[6] and 3D-techniques [7]. Single-shot projections are preferred in individuals who are unable to hold their breath, such as severely sick patients or small children [7]. 3D-imaging techniques provide better image quality compared to 2D-sequences [1,8,9], even though the combination of different MRCP sequences has proven to be valuable in the assessment of bile duct anatomy and pathology [10]. 3D FSE sequences are usually acquired with the slab in coronal orientation.

Maximum intensity projections (MIP) can then be obtained in any plane [7].Previous studies have addressed the matter of optimal slice thickness for data acquisition [11] and different techniques regarding respiratory triggering [12].However, to the best of our knowledge, the ideal acquisition and reconstruction plane, in practical terms meaning best suitable for optimal bile duct visualization with 3D techniques, has not been evaluated.The purpose of this study was to compare different acquisition and reconstruction planes of T2weighted 3D MRCP acquisitions for assessment of the intra-and extra-hepatic bile ducts.

Methods and materials
51 patients (30 female, 21 male, mean age 47.5 years, range 18-79 years) who were referred for liver MRI and dedicated MRCP were included in this prospective study, with approval of the institutional review board.

Inclusion criteria
Patient age equal to or greater than 18 years with suspected CBD pathologies

MR imaging technique
MR examinations were performed on a 3 Tesla system (Philips Ingenia) using dedicated multi-channel surface coils covering the abdomen.Prior to image acquisition, patients received 200 mL of a negative oral contrast agent for suppression of gastroenteric fluid signal.All patients underwent a clinical routine imaging protocol of the liver, including a respiratory-triggered 3D-MR cholangiography in the coronal (dataset A) as well as in the axial plane (dataset B) apart from 14 slices of 2D MRCP, single-shot breath-hold acquisition.Results from ERCP were considered as truth for sensitivity and specificity analysis.The specific MRCP sequences had sequential k-space filling with partial Fourier filling allowed, resulting in acquisition of central k-space lines approximately three minutes after the start of the MRCP sequence parameters are provided in detail in Table 1.

Image evaluation
Two readers independently performed image evaluation in terms of visibility of different bile duct segments up to the third order and assessment of technical quality.Readers were blinded to each patient's history and other imaging findings.A single coronal and axial maximum intensity projection (MIP) covering the central, left, right, and peripheral bile ducts was generated from each acquired MRCP dataset, resulting in two coronal and two axial MIP datasets, respectively.
Each reader evaluated the reconstructed MIP in the following way: 1, Coronal reconstructed MIP of the coronal acquisition vs. coronal reconstructed MIP of the axial acquisition; 2, Axial reconstructed MIP of the coronal acquisition vs. axial reconstructed MIP of the axial acquisition.
Depiction of bile duct segments was assessed using the following four-point scale proposed by Papanikolaou and colleagues [13]: 1, segment not seen; 2, segment faintly seen; 3, segment well seen but portion of the duct or the confluence not seen; and 4, excellent depiction including the proximal and distal portions.This scale was applied to the following sections (segments) of the biliary tract: the common bile duct (CBD), the right anterior bile duct, the right posterior bile duct, the left hepatic duct, and third-order biliary branches.
Overall technical image quality was assessed using a four-point scale proposed by Lim and colleagues [14]: 1, poor quality with severe artifacts; 2, satisfactory quality with few artifacts; 3, good quality with minimal artifacts; and 4, excellent quality without artifacts.
The two radiologists graded studies obtained with each sequence in a blinded fashion, and the paired student test was used to assess differences in technical quality and visibility of individual ductal segments of the biliary tree.

Statistical analysis
Results regarding bile duct visualization and overall technical image quality were compared with a two-sided Wilcoxon signedrank test after Bonferroni correction (with a p-value <0.05 deemed significant) in the following way:

Bile duct visualization
In case of coronal data acquisition, visualization of bile duct segments was significantly better on coronal reconstructed MIP as compared to axial reconstructed MIP (p < 0.05).This was true for visualization of the CBD, right anterior hepatic duct, left hepatic duct, and third-order biliary branches.In case of axial data acquisition, one reader observed a significantly better visualization of the CBD and left hepatic duct on coronal reconstructed MIP as compared to axial reconstructed MIP.
Regarding bile duct visualization, coronal MIP of the coronal acquisition (Dataset A) was equal to coronal MIP of the axial acquisition (Dataset B) (p > 0.05) (Figure 1).
Axial MIP of the axial acquisition (Dataset B) was significantly better than axial MIP of the coronal acquisition (Dataset A) for visualization of third-order biliary branches, whereas lower-order branches did not show a difference (Figures 2 and 3).
Bile duct visualization up to the third-order is equal on both datasets, even though the image impression is more blurred on the MIP derived from the axial acquired dataset (B).P values were calculated with the two-sided Wilcoxon Test after Bonferroni correction to compare depiction scores of coronal axial acquired datasets.

Technical image quality
Regarding overall technical image quality (including axial and coronal reconstructed MIP of a given dataset), there was no significant difference between the coronal and axial acquired datasets (p > 0.05).However, in the case of coronal data acquisition, detailed dataset analysis showed that technical image quality of the coronal MIP was significantly better as compared to the axial reconstructed MIP (p < 0.05).In the case of axial data acquisition, there was no significant difference regarding technical image quality of the reconstructed MIP (p > 0.05).

Choice of preferred image dataset
When reading coronal reconstructed MIP, readers preferred coronal acquisitions over axial acquisitions in 66% of the readings.Regarding axial MIP reconstruction, axial acquisitions were preferred over coronal acquisitions in 80% of the readings.Intraobserver agreement regarding choice of the preferred image dataset was excellent (weighted Κ range 0.94-1.00);interobserver agreement was moderate to excellent (weighted Κ range 0.57-0.85).
ERCP showed 30% malignant obstructions, 30% calculi in the common bile duct, 8% miscellaneous disorders and in 32% no abnormalities (Table 2).A significantly higher diagnostic accuracy of the 3D MRCP technique over the SS MRCP technique (p <0.05) using the McNemar's test was observed.Obstruction due to tumor was shown in 30% of patients, and calculi in the common bile duct were also shown in 30% of patients employing the 3D MRCP technique.Obstruction due to tumor and calculi were shown in 30% and 21% of patients, respectively, using the SS 2D MRCP technique.

ERCP 3D MRCP 2D MRCP
Malignant Obstruction 30% 30% 30% Calculi 30% 30% 21% Sensitivity and specificity in distinguishing calculi in the common bile duct by 3D MRCP coronal acquisition and SS MRCP were 100%, 100%, 70% and 100%, respectively.Interobserver agreement for 3D acquisition coronal MRCP was good for all diagnosis at a Kappa value ranging from 0.76 to 0.90, but bad to moderate for the SS MRCP at a Kappa value ranging from 0.20 to 0.63.

Discussion
To the best of our knowledge the ideal acquisition and reconstruction plane for optimal bile duct evaluation with 3D techniques has not yet been evaluated.For single-shot FSE techniques, it has been suggested that straight coronal and initial left posterior oblique images clearly depict the common hepatic duct and the left hepatic duct, whereas the CBD and right hepatic ducts are better seen in left posterior images obtained at a steeper angle [15].In a first approach towards projection cholangiography by means of MRI in 1991, Wallner and colleagues [4] used a heavily T2-weighted gradient echo sequence for assessment of bile duct dilatation.They concluded that imaging in the coronal plane provided a good view of the biliary system, whereas no additional information was found by imaging in the sagittal plane.In this study we compared different acquisition and reconstruction planes of T2weighted 3D MRCP acquisitions for assessment of the intra-and extra-hepatic bile ducts.In contrast to single-shot techniques, 3D MRCP has the advantage of facilitating secondary reconstructions.Coronal reconstructions were preferred, regardless of the initial acquisition plane.These findings were supported by good intra-and interobserver agreements.One of the reasons for coronal image preference may be the fact that these images are similar to image impressions of ERCP and conventional cholangiograms.
Other studies have evaluated secondary reconstruction techniques for MRCP.Schaible and colleagues [17] evaluated selective MIP reconstructions of respiratory-triggered 3D MRCP versus standard MIP reconstructions and single-shot MRCP.Single-shot and standard MIP reconstructions of 3D MRCP were comparable in terms of anatomical bile duct visualization, whereas selective MIP post-processing proved useful for detection of pathological alterations.In a retrospective study, Morita and colleagues [18] compared volume rendering (VR) and MIP of 3D-TSE MRCP sequences to define biliary anatomy mostly in patients without major biliary tract anomaly.Definition of biliary anatomy was found to be more accurate using VR reformation than MIP.However, the assessment of VR images was not the purpose of the present study.One disadvantage of VR reconstructions is that the detection degree of each structure depends on the setting of display parameters, particularly on the lower threshold of the opacity curve.Therefore, VR images need to be evaluated interactively [18].
In 1999, Boraschi and colleagues [19] compared axial and coronal 2D FSE sequences with 3D-MIP projection images in patients with suspected hepatobiliary disease.A higher global accuracy for axial and coronal FSE T2-weighted sequences was found regarding the diagnosis of the level and probable cause of biliary obstruction in depiction of small intraductal pathology, such as calculi or neoplastic lesions.
We have limited our analysis to reconstructed rather thin-slice source images, as the purpose of this specific study was to directly compare acquisition and reconstruction planes for MIP assessment.A well-known limitation of MIP is that small filling defects may be obscured due to partial volume effects [20].Further, overestimation of ductal narrowing and pseudostricture may result from the nature of MIP reconstruction [21].Therefore, it is important that MIP reconstructions not be appraised separately, but always in combination with the original acquired dataset, and in combination with other morphological sequences.

Conclusions
We compared different acquisition and reconstruction planes of T2-weighted 3D MRCP acquisitions for assessment of the intraand extrahepatic bile ducts in patients with different hepatobiliary pathologies.
The biggest disadvantage of 3D imaging is its acquisition time of five minutes, compared to the short time duration of 2D acquisition.
The results of our study suggest that coronal reconstructions are preferred for visualization and evaluation of the bile ducts,.In this context, the orientation of the primary dataset (coronal or axial) is negligible.
Although the 3D MRCP multislice technique is more timeconsuming than the SS MRCP breath-hold technique at a 3 Tesla (Philips 3 T Ingenia) system, it is advisable to use thin-slice 3D MRCP in order not to misdiagnose calculi in the common bile duct.Better interobserver agreement is reached employing the 3D MRCP