Secondary Logo

Share this article on:

Acceleration of Magnetic Resonance Cholangiopancreatography Using Compressed Sensing at 1.5 and 3 T: A Clinical Feasibility Study

Taron, Jana, MD*; Weiss, Jakob, MD*; Notohamiprodjo, Mike, MD*; Kuestner, Thomas, PhD; Bamberg, Fabian, MD*; Weiland, Elisabeth, PhD; Kuehn, Bernd, PhD; Martirosian, Petros, PhD§

doi: 10.1097/RLI.0000000000000489
Original Articles

Objectives Magnetic resonance cholangiopancreatography (MRCP) is an established technique in routine magnetic resonance examination. By applying the compressed sensing (CS) acceleration technique to conventional MRCP sequences, scan time can be markedly reduced. With promising results at 3 T, there is a necessity to evaluate the performance at 1.5 T due to wide scanner availabilities. Aim of this study is to test the feasibility of accelerated 3-dimensional (3D) MRCP with extended sampling perfection with application-optimized contrasts using different flip angle evolution (SPACE) using CS in navigator triggering and in a single breath-hold in a clinical setting at 1.5 T and 3 T and compare it with a conventional navigator-triggered 3D SPACE-MRCP.

Materials and Methods Phantom measurements were performed to adapt sequence parameters. Conventional 3D SPACE-MRCP in navigator triggering (STD_MRCP) as well as CS-accelerated 3D SPACE-MRCP acquired in navigator triggering and in a single breath-hold (CS_MRCP and CS_BH_MRCP) was performed in 66 patients undergoing clinically induced MRI of the pancreatobiliary system at 1.5 T and 3 T. Image quality evaluation was performed by 2 independent radiologists. Dedicated statistics were performed (P < 0.05 considered significant).

Results In patient imaging, CS_MRCP was superior to STD_MRCP and CS_BH_MRCP in aspects of overall image quality at 1.5 T (P = 0.01; P < 0.001) and 3 T (P = 0.002; P = 0.013). Overall image quality in CS_BH_MRCP was inferior compared with STD_MRCP and CS_MRCP at 1.5 T. At 3 T, overall image quality in CS_BH_MRCP was superior to STD_MRCP (P = 0.001). Scan time was reduced by 25% to 46% covering 5% of k-space (CS_MRCP at 1.5 and 3 T) and 97% covering 3.6% of k-space (CS_BH_MRCP at 1.5 and 3 T).

Conclusions Compressed sensing-accelerated MRCP is feasible in clinical routine at 1.5 and 3 T offering major reduction of acquisition time. When applying a single breath-hold CS imaging, field strengths of 3 T are recommended.

From the *Department of Diagnostic and Interventional Radiology, University Hospital of Tuebingen, Tuebingen;

Institute of Signal Processing and System Theory, University of Stuttgart, Stuttgart;

Siemens Healthcare GmbH, Erlangen; and

§Section on Experimental Radiology, Department of Diagnostic and Interventional Radiology, University Hospital of Tuebingen, Tuebingen, Germany.

Received for publication March 6, 2018; and accepted for publication, after revision, April 26, 2018.

Conflicts of interest and sources of funding: none declared.

Supplemental digital contents are available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.investigativeradiology.com).

Correspondence to: Jana Taron, MD, Department of Diagnostic and Interventional Radiology, University Hospital of Tuebingen, Hoppe-Seyler-Str. 3 72076 Tuebingen, Germany. E-mail: jana.taron@med.uni-tuebingen.de.

Magnetic resonance cholangiopancreatography (MRCP) was first introduced in the 1990s and has since become an established imaging technique for noninvasive examination of the biliary tract.1–3 Although endoscopic retrograde cholangiopancreatography is still considered the reference standard for imaging of the pancreatic and biliary ductal system,2,4,5 MRCP has proven useful due to noninvasiveness, comparable accuracy and cost-savings, especially when endoscopic retrograde cholangiopancreatography is problematic or nonapplicable.2,5–7 Standard MRCP typically uses T2-weighted 3-dimensional turbo spin echo (3D TSE) sequences,8,9 resulting in images with a high contrast between the bile fluid and the surrounding parenchyma.2 Further optimization can be achieved by applying a 3D MRCP using a sampling perfection with application-optimized contrasts (SPACE) sequence8,10,11 and respiratory triggering to reduce respiratory motion artifacts, which is done in a prospective navigator-triggered acquisition correction (PACE) scouting the liver dome and monitoring its movements.9,12 This MRCP protocol consisting of a 3D navigator-triggered SPACE is widely established in clinical routine requiring an average of 3 to 7 minutes of acquisition time depending on the patients breathing pattern.11

Based on the theory of compressed sensing (CS), it has lately been shown that sparsity-based magnetic resonance imaging (MRI) scans can be used to significantly reduce scan time in MRI.13–16 The CS algorithms are used to reconstruct images from a small percentage of randomly acquired k-space data demanding the following characteristics: (a) random, sparse, and incoherent subsampling (meaning the final image having a sparse representation in a particular transform domain with the randomly acquired subsampling artifacts being incoherent) and (b) a nonlinear iterative reconstruction (to reduce subsampling-related aliasing artifacts).13–15

Applying the CS acceleration technique to the conventional 3D SPACE sequence is a promising step toward reduction of acquisition time in imaging of the pancreatobiliary system. The feasibility of CS-accelerated MRCP has been tested and proven in previous studies at field strengths of 3 T suggesting higher resolution and a notable scan time reduction compared with standard 3D MRCP.13,16–18 Although these results are convincing at 3 T, there is a paucity of investigations on the performance of CS-accelerated MRCP at 1.5 T. The generally lower signal-to-noise ratio (SNR) at 1.5 T—compared with images obtained at 3 T—could be one challenging factor in the application of CS. Combined with the subsampling of k-space in CS, which also reduces image signal, SNR might be drastically reduced potentially yielding to poor image quality of nondiagnostic value. In literature, the comparison of conventional 3D MRCP images at 1.5 and 3 T have led to superior results in image quality in the 3 T approach.11,19,20 Yet, this is an important aspect that needs to be addressed as field strengths of 1.5 T are widely spread in clinical routine and 3 T is not always available.

A second major aspect in MR examinations are breathing-related artifacts. Patients of senior age or in critical conditions are less likely to tolerate long-lasting examinations, others might present irregular breathing cycles, which may interfere with standard navigator-triggered protocols. To address this problem, a single breath-hold protocol is highly desired and can be achieved by applying CS.

Thus, we aim to compare the performance of a CS-accelerated MRCP to a conventional MRCP in clinical routine imaging of the pancreatobiliary tract at 1.5 T and 3 T. In a first step, phantom measurements are used to adapt sequence parameters and to eliminate potential patient-related artifacts. In a second step, the CS-accelerated MRCP is obtained with 2 different breathing regimes (navigator triggering and single breath-hold) in a clinical setting.

Back to Top | Article Outline

MATERIALS AND METHODS

This study was approved by the local institutional review board and written informed consent for participation and scientific data evaluation was waived.

In a first step, the CS-accelerated MRCP as well as the conventional MRCP sequences were applied on a biliary-structure phantom especially built for this project to adapt sequence parameters and ensure optimal image quality. In a second step, sequence parameters were transferred to clinical routine protocols for patients scheduled for clinically indicated MRI of the pancreatobiliary system.

Patient data and information were drawn from electronic medical records.

Back to Top | Article Outline

Phantom

For phantom measurements, 6 different angiographic catheters with outer diameters of 3.3F, 4F, 5F, 8F, 12F, and 16F (1F ≈ 0.333 mm) were chosen. Under the assumption that a 3F catheter has an inner diameter of approximately 2F, this corresponds to the following inner diameters of 0.7 mm, 0.9 mm, 1.1 mm, 1.8 mm, 2.7 mm, and 3.6 mm.21 These catheters were filled with 0.9% saline solution and placed in an acrylic glass container filled with gadolinium-doped distilled water (≈1 mM) to suppress background enhancement.

Back to Top | Article Outline

Patients

Between March and July 2017, patients scheduled for clinically indicated MRCP were included in this study. Standard examination preparation (in absence of contraindications) for each patient included 200 mL of pineapple juice orally, functioning as negative contrast agent,22 as well as 1 ampule (20 mg/mL) of butylscopolaminiumbromid intravenously to reduce artifacts caused by bowel motion.

Back to Top | Article Outline

MRI Protocol

Examinations were performed on 1.5 T (MAGNETOM Aera and Avantofit; Siemens Healthcare, Erlangen, Germany) and 3 T (MAGNETOM Skyra and Prismafit; Siemens Healthcare, Erlangen, Germany) whole-body MRI scanners using an 18-channel body coil combined with a 32-channel spine coil. For imaging of the biliary tract, a conventional 3D SPACE as well as a prototype 3D CS-SPACE sequence provided by Siemens (Siemens Healthcare, Erlangen, Germany) were used. The CS-SPACE sequence uses a Poisson-Disk variable density subsampling pattern, which offers increased incoherent sample statistics and reduced subsampling artifacts. Acceleration in the CS-SPACE sequence is specified by the sampling factor (SF), which is the ratio of the sampled k-space data to the entire k-space. In the reconstruction step, the specific features of the prototype sequence are nonlinear iterative reconstruction integrating SENSE15 and L1 wavelet regularization in phase and slice encoding directions.23

In vivo, the conventional SPACE sequence was performed with navigator triggering (STD_MRCP); the accelerated sequences were conducted using navigator triggering (CS_MRCP) as well as a single breath-hold examination (CS_BH_MRCP). In phantom measurements, no trigger was used. Instead, a repetition time (TR) of 4000 milliseconds was chosen to approximate the breathing cycle. Acceleration was achieved due to an SF of 0.036 in CS_BH_MRCP and 0.05 in CS_MRCP, thus acquiring 3.6% and 5% of the entire k-space, respectively. In CS_BH_MRCP, shorter echo time was chosen to increase SNR. In addition, to compensate for the reduction of SNR, the resolution of CS_BH_MRCP was slightly reduced.

Coronal slice selection was applied. In patients, phase-encoding direction was chosen right-left; in phantom measurements phase-encoding direction was chosen feet-head as well as right-left (ie, parallel and orthogonal to the catheters).

For detailed sequence parameters for in vivo measurements, see Table 1. Detailed sequence parameters for phantom measurements are available in Supplemental Digital Content 1, http://links.lww.com/RLI/A385.

TABLE 1

TABLE 1

Back to Top | Article Outline

Image Quality Assessment

Qualitative image evaluation was performed independently and in a blinded fashion by 2 radiologists (J.T. and J.W.) with 4 and 3 years of experience in MRI. The images were assessed subsequently on a 5-point Likert scale according to the following criteria: (1) overall image quality (including breathing-, motion-, and CS-related artifacts), (2) visibility of intrahepatic ducts, (3) visibility of common bile duct, (4) visibility of pancreatic duct, (5) sharpness of ducts, (6) lesion conspicuity, and (7) sharpness of lesion. The 5-point Likert scale was defined as follows: overall image quality (5 excellent, 4 good, 3 moderate, 2 poor, 1 nondiagnostic), visibility of ducts or lesions (5 excellent visibility, 4 good visibility, 3 moderate visibility, 2 poor visibility, 1 not visible), and sharpness of ducts and lesions (5 excellent sharpness, 4 minimal blur, 3 moderate blur, 2 significant blur, 1 nondiagnostic).

Lesions were defined as cysts, stenosis, intraluminal signal loss, or duct irregularities. Standard of reference for lesions was based on MRI scans consisting of the anatomical sequences included in the contrast-enhanced pancreatobiliary imaging protocol (STD-MRCP included). Multiple lesions per patient were pooled in the assessment.

Image acquisition time was assessed.

Back to Top | Article Outline

Statistical Analysis

Statistical analysis was conducted with a dedicated software (JMP 11.0; SAS Institute, Cary, NC; SPSS Version 23.0; IBM Corporation, Chicago, IL). Qualitative image parameters were compared between sequences using pairwise Dunn-Bonferroni post hoc test in case of significant difference in the Wilcoxon signed-rank test. P < 0.05 was considered statistically significant.

Interobserver agreement was calculated using Cohen kappa coefficient (<0.2 poor, 0.2 to 0.4 fair, 0.41 to 0.6 moderate, −0.61 to 0.8 good, and 0.81 to 1.0 perfect agreement).24

Back to Top | Article Outline

RESULTS

Phantom Study

Results of the phantom study are shown in Figures 1 and 2. Image quality did not differ when comparing images acquired at 1.5 T and 3 T. STD_MRCP and CS_MRCP demonstrated equivalently high quality in displaying all catheter structures of the phantom. CS_BH_MRCP presented inferior image quality with reduced signal and sharpness of structures; catheters smaller than 8F were not appreciated.

FIGURE 1

FIGURE 1

FIGURE 2

FIGURE 2

Gibbs artifacts were apparent when choosing phase encoding in right-left direction, pronounced in CS_MRCP at 3 T. Stair-step artifacts were noticed, most prominent in structures smaller than 8F in images obtained with CS_MRCP at 1.5 T.

Acquisition time in the conventional MRCP was 5:49 minutes; in CS_MRCP with an SF of 0.05, it was 2:56 minutes; and in CS_BH_MRCP with an SF of 0.036 scan time, it was 13 seconds.

Back to Top | Article Outline

Patients Study

Sixty-six patients were included in this study (31 women, 35 men; mean age, 45.7 ± 7.5 years), 46 patients received imaging at 1.5 T, 20 received imaging at 3 T. Indications for MRCP were the following: n = 19 follow-up of suspected intraductal papillary mucinous neoplasm, n = 9 further diagnostics of sonographically detected pancreatic lesion, n = 7 follow-up examination of cystic pancreatic lesion, n = 5 suspected cholangiocellular carcinoma, n = 6 primary sclerosing cholangitis, n = 5 follow-up after liver transplant, n = 4 follow-up after surgically removed cholangiocellular carcinoma, n = 4 follow-up after hemihepatectomy, n = 3 suspected choledocholithiasis, n = 2 suspected cholangitis, and n = 2 suspected autoimmune pancreatitis.

Results of the qualitative evaluation are displayed in Table 2.

TABLE 2

TABLE 2

At 1.5 T, CS_MRCP was found superior to STD_MRCP in overall image quality (P = 0.001), visibility of pancreatobiliary structures (common bile duct P = 0.001, pancreatic duct P = 0.01), lesion conspicuity (P = 0.002), and lesion sharpness (P = 0.008). Statistical significance was not reached in visibility of intrahepatic ducts. CS_BH_MRCP was rated lower compared with STD_MRCP; statistical significance was reached in visibility of intrahepatic ducts (P = 0.008) and sharpness of ducts (P = 0.03). When comparing CS_MRCP to CS_BH_MRCP, the single breath-hold sequence resulted in significantly inferior ratings in all aspects (each P < 0.001). Representative images at 1.5 T are displayed in Figures 3 and 4 (Supplemental Digital Contents 2 to 4 for Figure 4, Video File 1 [Video file of all slices of original data acquired in STD_MRCP at 1.5 T], http://links.lww.com/RLI/A386; Video File 2 [Video file of all slices of original data acquired in CS_MRCP at 1.5 T], http://links.lww.com/RLI/A387; and Video File 3 [Video file of all slices of original data acquired in STD_MRCP at 1.5 T], http://links.lww.com/RLI/A388).

FIGURE 3

FIGURE 3

FIGURE 4

FIGURE 4

At 3 T, CS_MRCP was rated significantly superior to STD_MRCP in all image quality aspects except intrahepatic ducts (overall image quality P = 0.002, visibility of common bile duct and pancreatic duct P = 0.002, duct sharpness P = 0.009, lesion conspicuity P = 0.005, lesion sharpness P = 0.001). CS_BH_MRCP was found equivalent to superior to STD_MRCP reaching statistical significance in overall image quality (P = 0.001), visibility of the pancreatic duct (P = 0.001), sharpness of ducts (P = 0.021), and sharpness of lesions (P = 0.043). In the comparison of CS_MRCP to CS_BH_MRCP, the single breath-hold sequence resulted in lower image quality reaching a statistically significant difference in overall image quality (P = 0.013) and visibility of the common bile duct (P = 0.002). Representative images at 3 T are displayed in Figures 5 and 6 (Supplemental Digital Contents 5 to 7 for Figure 5, Video File 4 [Video file of all slices of original data acquired in STD_MRCP at 3 T], http://links.lww.com/RLI/A389; Video File 5 [Video file of all slices of original data acquired in CS_MRCP at 3 T], http://links.lww.com/RLI/A390; and Video File 6 [Video file of all slices of original data acquired in CS_BH_MRCP 3 T], http://links.lww.com/RLI/A391).

FIGURE 5

FIGURE 5

FIGURE 6

FIGURE 6

When comparing image quality ratings at different magnetic field strengths, 3 T delivered higher ratings in image quality, higher visibility of the different portions of the pancreatobiliary system, and lesions. Sharpness of ducts and lesions was rated equivalent to superior.

Interobserver agreement was moderate for STD_MRCP and CS_MRCP (with 0.592 and 0.566, respectively) and good for CS_BH_MRCP with 0.616.

Mean acquisition times at 1.5 T were the following: STD_MRCP 6:39 minutes (range, 4:24–9:19 minutes), CS_MRCP 5:01 minutes (range, 3:05–7:37 minutes), and CS_BH_MRCP 13 seconds. Mean acquisition time at 3 T was: STD_MRCP 7:26 minutes (range, 4:16–11:15 minutes), CS_MRCP 4:00 minutes (range, 2:48–7:14 minutes), and CS_BH_MRCP 13 seconds. Thus, when acquiring 5% of k-space (CS_MRCP) a scan time reduction of 25% at 1.5 T and 46% at 3 T was achieved, when acquiring 3.6% of k-space (CS_BH_MRCP) a reduction of scan time of 96.7% and 97.1% (at 1.5 T and at 3 T) was achieved compared with the STD_MRCP approach.

Back to Top | Article Outline

DISCUSSION

Our results demonstrate that the CS-accelerated MRCP is feasible at field strengths of 1.5 T and 3 T in phantom studies and in clinical routine providing good to excellent image quality and a considerable scan time reduction.

Depending on the breathing scheme and field strength, scan time can be reduced to 54% to 75% of the mean time in standard acquisition covering 5% of k-space using a navigator triggering approach and to 3% covering 3.6% of k-space in a single breath-hold. Previously performed studies investigating CS-accelerated MRCP with different breathing schemes at 3 T demonstrated similar results. Yoon et al described a decrease in mean acquisition time from 419 seconds in the conventional MRCP to 227 seconds in the CS-approach (approximately 54% of the initial mean scan time) by sampling 4.5% to 5.7% of k-space data.18 Zhu and colleagues covered 4.2% to 4.5% of k-space data in their investigation, which resulted in a scan time of 17 seconds in the single breath-hold examination and a mean scan time of 134 seconds in the navigator-triggered CS-MRCP (approximately 3% and 37% of the mean scan time of 364 seconds in the standard MRCP).17

Although these are drastic reductions in examination time, it nevertheless needs to be mentioned that these results can not only be accounted for by CS acceleration but also by reduced spatial resolution in the case of CS_BH_MRCP (1.0 × 1.0 × 1.0 vs 1.2 × 1.5 × 2.4 mm3). Furthermore, scan time in CS_BH_MRCP is determined by the fixed TR, whereas scan time in navigator-triggered sequences depends on the patient's breathing cycle (which is usually markedly longer compared with the TR of 1700 milliseconds in the breath-hold examination).

When comparing conventional and CS-accelerated images in phantom measurements obtained at different field strengths at 1.5 T and 3 T, no differences in image quality were noted. Image quality achieved by STD_MRCP and CS_MRCP was found equivalent, whereas the CS_BH_MRCP approach demonstrated markedly reduced image quality compared with the aforementioned with reduced signal and edge sharpness. Signal reduction in this approach resulted in loss of visibility of catheter structures finer than 16F/8F, respectively, which can be attributed to the massive subsampling of k-space and the missing of high-frequency components. The subsampling, especially in the periphery of k-space, leads to a low-pass filter effect and, thus, to loss of resolution of the final image. These effects have previously been described in patient imaging at 3 T by Zhu at al17 and, likewise, sustain in the images of our patient study. We noticed an underrepresentation of fine or peripherally located structures of the pancreatobiliary tract in CS_BH_MRCP, an important phenomenon to keep in mind when reading these images, as it might be misinterpreted as a lesion, irregularity, or intraluminal structure.

In patient imaging, CS_MRCP proved to be the superior sequence regardless of the applied field strength. CS_BH_MRCP delivered only moderate image quality at 1.5 T with inferior results compared with CS_MRCP and the conventional sequence. At 3 T, on the other hand, image quality of the single breath-hold sequence improved presenting comparable to CS_MRCP and superior to STD_MRCP.

There are various aspects affecting image quality that need to be discussed in this context. One important point is the previously described signal loss and loss of image sharpness at 1.5 T. In return, higher signal and sharper edges resulting from higher field strengths contribute to the improved image quality at 3 T. Besides these technical aspects, breathing- and motion-related artifacts are a further major factor modulating image quality. Irregularities in breathing pattern can cause a reduction of image quality in the navigator-triggered MRCP sequences (standard and CS-accelerated) leading to a superiority of the CS_BH_MRCP despite lower SNR, which was shown in previous literature.13,17,18 Especially in the critically ill or those of advanced age with limited tolerance to long examination times, this is a major and not to be neglected aspect closely associated to the resulting image quality. Although this could suggest a superiority of CS_BH_MRCP in patient imaging, the results in our study demonstrated a higher mean overall image quality in the navigator-triggered CS sequence as opposed to the breath-hold sequence. We believe that one contributing factor to these results is the patient collective itself. Given the relatively young age of our study collective (with a mean of 45.7 ± 7.5 years), these patients might have been more cooperative during MR examination and potentially presented a more regular breathing pattern. Taking this into account, the reduction of SNR based on the coverage of a lower percentage of k-space (as present in CS_BH_MRCP) seems to have had a greater impact on image quality than the benefits of a breath-hold sequence in our patient population. Yet, the higher resistance to motion-related artifacts due to faster image acquisition can still be accounted for the superiority of the navigator-triggered CS-accelerated approach when compared with the conventional sequence.

Although our findings at 3 T are consistent with recent literature reports in which CS-accelerated MRCP in navigator triggering delivered superior image quality compared with conventional sequences and CS-accelerated breath-hold protocols,17,18 comparable studies at 1.5 T have not yet been published. Yet, with a lack of 3 T scanners in many institutions, but—on the other hand—a widespread availability of 1.5 T scanners, we find this result of high relevance in future patient care. Although we generally favor the navigator-triggered CS-approach in diagnostic imaging, we see the application of the CS breath-hold sequence elsewhere. Despite potential underrepresentation of fine structures or reduced image signal, this protocol can deliver important information, especially when imaging of the pancreatobiliary system is urgently needed for further therapeutic decisions and patient endurance does not allow longer scan times. In this respect, if a certain reduction of image quality is acceptable, CS_BH_MRCP is an alternative offering major scan time reduction at both field strengths.

Surprisingly, the discrepant results of image quality of the CS_BH_MRCP sequence in phantom (with significantly reduced image quality) compared with in vivo measurements (with moderate to very good image quality) need to be mentioned. A potential explanation could be the structural shape and strictly parallel alignment of catheter tubes in the phantom. Depending on their orientation (orthogonal vs parallel to the phase encoding direction), the potency of spatial sparsity can be exploited differently as well as reconstruction constraints have distinct contribution.

Other important findings that need to be addressed are CS-associated artifacts, which are a common sighting and have widely been discussed in literature.25–27 The most commonly encountered artifacts are those of global ringing, which can appear like motion artifacts and blurring.16 In our study, we also discovered the appearance of ringing as well as stair-step artifacts, which were observed best in our phantom studies, to a much lower extent in patient imaging.

Ringing artifacts (also called Gibbs artifacts) are represented by multiple parallel lines immediately adjacent to high-contrast interfaces. These artifacts are caused during Fourier transformation when the signal is transformed to the image by false widening or distortions of the edges at interfaces. As Gibbs artifacts are a consequence of the Fourier transformation, they occur in phase and frequency direction but are most pronounced in phase encoding due to fewer samples.28 This is nicely demonstrated in our phantom images. With the phase-encoding direction feet-head, artifacts are markedly reduced as catheters run in the same way. Whereas, when the phase-encoding direction is chosen right-left, ringing at interfaces increases due to the increasing number of interfaces in this direction (as shown in Figs. 1, 2). These findings can be transferred to our experiences in patient imaging. The main directions of the bile duct are mostly in coronal plane running from feet to head (which was chosen as phase-encoding direction). Yet, bile ducts do not strictly obey directions. Especially, the pancreatic ducts run almost perpendicular to this plane, which appears less sharp and at times widened. Furthermore, this could explain the reduction of peripheral ducts as these disperse into the parenchyma. If a structure runs perpendicular to the phase-encoding direction, it appears wider and less sharp, which is the result of reconstruction algorithm, filter, and subsampling. A recommendation to overcome these artifacts is to adapt the phase-encoding direction accordingly, if the structures desired to image run in a preferred direction.

Stair-step artifacts were the second artifact that we faced in our examinations. These were visible in finer structures and more pronounced in the CS-accelerated images than in STD_MRCP. This artifact is related to the in-plane resolution of the image. In the conventional sequence, acceleration is achieved through parallel imaging in in-plane only. In the CS sequence, slice- and phase-encoding directions are subsampled bringing along a reduced spatial resolution, which results in these wave-like edge irregularities. Furthermore, we believe that the postprocessed maximum intensity projections (such as displayed in our images) aggravate the impression of these artifacts. Interestingly, stair-step artifacts were hardly discovered in patient images. One potential explanation might lie within the reconstruction algorithm itself, which is optimized for in vivo data and might account for this problem.

There are limitations to this study. This retrospective study includes only a limited number of patients that were examined at 1.5 T or 3 T, respectively, not allowing a direct comparison of images acquired at different field strengths. Yet, the implication of this study to test the feasibility in clinical application was achieved. Second, to compensate the lower SNR, the resolution of the CS_BH_MRCP sequence was reduced and higher refocusing flip angles were possible at 1.5 T. Further modification of parameters to improve image quality at 1.5 T would have been possible, but to allow direct comparison the remaining imaging parameters were kept identical. Third, as this was a prototype sequence, the percentage of k-space acquired was chosen according to vendor recommendations and kept identical throughout the study to allow comparison. Nevertheless, it should be kept in mind that there might be room for optimization of k-space sampling, which could further optimize this imaging approach. Furthermore, artifacts were not evaluated separately, but evaluation of imaging artifacts was subsumed under the aspect of “overall image quality.” Moreover, reference standard for presence and character of a lesion detected in the sequences were the morphological MR sequences. As most of the patients demonstrated lesions of benign appearance or those suggesting monitoring, a histopathological confirmation was not available in the majority of cases. Nevertheless, the focus of this evaluation was on image quality and detection of lesions in patient imaging, lesion characterization was beyond the scope of this study.

Concluding, CS-accelerated MRCP is feasible in clinical routine regardless of the applied field strength offering a major reduction of acquisition time. CS_MRCP is the preferred sequence for imaging the pancreatobiliary system in terms of image quality and acquisition time at 1.5 T and 3 T. In patients with irregular breathing pattern—and thus malfunction of the navigator triggering—or ill patients with limited endurance to lengthy examinations, CS_BH_MRCP can be of major help. Based on our experiences, in these cases imaging at 3 T should be favored. Yet, misinterpretation of biliary structures due to the CS algorithm, especially when less than 5% of k-space are sampled, should be kept in mind.

Back to Top | Article Outline

ACKNOWLEDGMENT

The authors would like to thank Matthias Kündel and Andreas Lingg for their great support in data acquisition.

Back to Top | Article Outline

REFERENCES

1. Sung S, Jeon TY, Yoo SY, et al. Incremental value of MR cholangiopancreatography in diagnosis of biliary atresia. PLoS One. 2016;11:e0158132.
2. Al-Dhuhli H. Role of magnetic resonance cholangiopancreatography in the evaluation of biliary disease. Sultan Qaboos Univ Med J. 2009;9:341–352.
3. Katabathina VS, Dasyam AK, Dasyam N, et al. Adult bile duct strictures: role of MR imaging and MR cholangiopancreatography in characterization. Radiographics. 2014;34:565–586.
4. Drake LM, Anis M, Lawrence C. Accuracy of magnetic resonance cholangiopancreatography in identifying pancreatic duct disruption. J Clin Gastroenterol. 2012;46:696–699.
5. Meagher S, Yusoff I, Kennedy W, et al. The roles of magnetic resonance and endoscopic retrograde cholangiopancreatography (MRCP and ERCP) in the diagnosis of patients with suspected sclerosing cholangitis: a cost-effectiveness analysis. Endoscopy. 2007;39:222–228.
6. Shanmugam V, Beattie GC, Yule SR, et al. Is magnetic resonance cholangiopancreatography the new gold standard in biliary imaging? Br J Radiol. 2005;78:888–893.
7. Howard K, Lord SJ, Speer A, et al. Value of magnetic resonance cholangiopancreatography in the diagnosis of biliary abnormalities in postcholecystectomy patients: a probabilistic cost-effectiveness analysis of diagnostic strategies. Int J Technol Assess Health Care. 2006;22:109–118.
8. Yoon LS, Catalano OA, Fritz S, et al. Another dimension in magnetic resonance cholangiopancreatography: comparison of 2- and 3-dimensional magnetic resonance cholangiopancreatography for the evaluation of intraductal papillary mucinous neoplasm of the pancreas. J Comput Assist Tomogr. 2009;33:363–368.
9. Morita S, Ueno E, Suzuki K, et al. Navigator-triggered prospective acquisition correction (PACE) technique vs. conventional respiratory-triggered technique for free-breathing 3D MRCP: an initial prospective comparative study using healthy volunteers. J Magn Reson Imaging. 2008;28:673–677.
10. Nakaura T, Kidoh M, Maruyama N, et al. Usefulness of the SPACE pulse sequence at 1.5 T MR cholangiography: comparison of image quality and image acquisition time with conventional 3D-TSE sequence. J Magn Reson Imaging. 2013;38:1014–1019.
11. Arizono S, Isoda H, Maetani YS, et al. High-spatial-resolution three-dimensional MR cholangiography using a high-sampling-efficiency technique (SPACE) at 3 T: Comparison with the conventional constant flip angle sequence in healthy volunteers. J Magn Reson Imaging. 2008;28:685–690.
12. Asbach P, Klessen C, Kroencke TJ, et al. Magnetic resonance cholangiopancreatography using a free-breathing T2-weighted turbo spin-echo sequence with navigator-triggered prospective acquisition correction. Magn Reson Imaging. 2005;23:939–945.
13. Chandarana H, Doshi AM, Shanbhogue A, et al. Three-dimensional MR cholangiopancreatography in a breath hold with sparsity-based reconstruction of highly undersampled data. Radiology. 2016;280:585–594.
14. Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn Reson Med. 2007;58:1182–1195.
15. Beck A, Teboulle M. A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM Journal on Imaging Sciences. 2009;2:183–202.
16. Seo N, Park MS, Han K, et al. Feasibility of 3D navigator-triggered magnetic resonance cholangiopancreatography with combined parallel imaging and compressed sensing reconstruction at 3 T. J Magn Reson Imaging. 2017;46:1289–1297.
17. Zhu L, Wu X, Sun Z, et al. Compressed-sensing accelerated 3-dimensional magnetic resonance cholangiopancreatography: application in suspected pancreatic diseases. Investigative Radiology. 2018;53:150–157.
18. Yoon JH, Lee SM, Kang HJ, et al. Clinical feasibility of 3-dimensional magnetic resonance cholangiopancreatography using compressed sensing: comparison of image quality and diagnostic performance. Invest Radiol. 2017;52:612–619.
19. Isoda H, Kataoka M, Maetani Y, et al. MRCP imaging at 3.0 T vs. 1.5 T: preliminary experience in healthy volunteers. J Magn Reson Imaging. 2007;25:1000–1006.
20. Onishi H, Kim T, Hori M, et al. MR cholangiopancreatography at 3.0 T: intraindividual comparative study with MR cholangiopancreatography at 1.5 T for clinical patients. Invest Radiol. 2009;44:559–565.
21. Iserson KV. J.-F.-B. Charrière: the man behind the "French" gauge. J Emerg Med. 1987;5:545–548.
22. Riordan RD, Khonsari M, Jeffries J, et al. Pineapple juice as a negative oral contrast agent in magnetic resonance cholangiopancreatography: a preliminary evaluation. Br J Radiol. 2004;77:991–999.
23. Liu J, Rapin J, Chang TC, et al. Dynamic cardiac MRI reconstruction with weighted redundant Haar wavelets. In: Proceedings of the 20th Annual Meeting of ISMRM, Melbourne, Australia, May 2012:178.
24. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159–174.
25. Worters PW, Sung K, Stevens KJ, et al. Compressed-sensing multispectral imaging of the postoperative spine. J Magn Reson Imaging. 2013;37:243–248.
26. Sharma SD, Fong CL, Tzung BS, et al. Clinical image quality assessment of accelerated magnetic resonance neuroimaging using compressed sensing. Invest Radiol. 2013;48:638–645.
27. Zhang T, Chowdhury S, Lustig M, et al. Clinical performance of contrast enhanced abdominal pediatric MRI with fast combined parallel imaging compressed sensing reconstruction. J Magn Reson Imaging. 2014;40:13–25.
28. Block KT, Uecker M, Frahm J. Suppression of MRI truncation artifacts using total variation constrained data extrapolation. Int J Biomed Imaging. 2008;2008:184123.
Keywords:

MRCP; compressed sensing; 1.5 T; 3 T; phantom

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.