Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Aug 1.
Published in final edited form as: Magn Reson Med. 2013 Sep 9;72(2):477–484. doi: 10.1002/mrm.24939

Reproducibility Study of 4D Flow MRI of Arterial and Portal Venous Liver Hemodynamics: Influence of spatio-temporal resolution

Zoran Stankovic 1,2, Bernd Jung 2, Jeremy Collins 1, Maximilian F Russe 2, James Carr 1, Wulf Euringer 2, Lena Stehlin 2, Zoltan Csatari 2, Peter C Strohm 3, Mathias Langer 2, Michael Markl 1
PMCID: PMC3950306  NIHMSID: NIHMS525862  PMID: 24018798

Abstract

Purpose

To evaluate influence of variation in spatio-temporal resolution and scan-rescan reproducibility on 3D visualization and quantification of arterial and portal venous (PV) liver hemodynamics at 4D flow MRI.

Methods

Scan-rescan reproducibility of 3D hemodynamic analysis of the liver was evaluated in 10 healthy volunteers using 4D flow MRI at 3T with three different spatio-temporal resolutions (2.4×2.0×2.4mm3, 61.2ms; 2.5×2.0×2.4mm3, 81.6ms; 2.6×2.5×2.6mm3, 80ms) and thus different total scan times. Qualitative flow analysis used 3D streamlines and time-resolved particle traces. Quantitative evaluation was based on maximum and mean velocities, flow volume and vessel lumen area in the hepatic arterial and PV systems.

Results

4D flow MRI showed good inter-observer variability for assessment of arterial and PV liver hemodynamics. 3D flow visualization revealed limitations for the left intrahepatic PV branch. Lower spatio-temporal resolution resulted in underestimation of arterial velocities (mean 15%, p<0.05). For the PV system hemodynamic analyses showed significant differences in the velocities for intrahepatic portal vein vessels (p<0.05). Scan-rescan reproducibility was good except for flow volumes in the arterial system.

Conclusion

4D flow MRI for assessment of liver hemodynamics can be performed with low interobserver variability and good reproducibility. Higher spatio-temporal resolution is necessary for complete assessment of the hepatic blood flow required for clinical applications.

Keywords: 4D flow MRI, liver hemodynamics, splanchnic system, reproducibility study, spatiotemporal resolution

Introduction

Patients with liver cirrhosis develop portal hypertension with attendant alterations in the liver hemodynamics. This results in a hyperdynamic syndrome with elevated cardiac output, increased portal blood flow and an elevated portosystemic pressure gradient (1). In advanced liver cirrhosis, portal blood flow represents the primary driver of elevated portal venous pressures. Consequently, pharmacological and interventional therapies minimize or redirect splanchnic blood flow in order to release portal venous pressure (2,3). Subsequently it is of interest in the clinic to measure the portal venous blood flow including the flow rate and maximum flow velocity as well as qualitatively assess flow patterns to non-invasively determine the response to therapy.

Time-resolved (cine) phase contrast (PC) MRI is regularly used for hemodynamic assessment of the thoracic (4) and abdominal aorta (5). 4D flow MRI representing time-resolved phase contrast (PC) MRI with tri-directional velocity encoding and three-dimensional (3D) anatomic coverage has recently gained increased attention due to its ability to provide comprehensive 3D flow visualization and quantitative blood flow evaluation (6–16). Previous reports include the application of 4D flow MRI in the thoracic aorta (6,7), pulmonary artery (8,9), and carotid artery (10), which represent high flow arterial systems. Recently, several studies qualitatively and quantitatively evaluated 3D blood flow in the splanchnic vessels including the low flow portal venous system using either a non-contrast enhanced phase contrast sequence (12,14,16) or a radial under sampling technique (13,15).

Previous studies comparing 4D flow MRI to standard 2D PC MRI and Doppler US indicated the validity of 4D flow based flow quantification in the aorta and surrounding great vessels (17,18), the carotid arteries (10) and recently for the portal venous system (12–16). Good reproducibility and low variability was demonstrated for 4D flow MRI in the aorta (19) with improved performance at 3T compared to 1.5T (20). However, a detailed evaluation of scan-rescan reliability and influence of spatiotemporal resolution in the low flow hepatic portal venous system as well as hepatic arterial system is lacking.

The purpose of this study was therefore to systematically evaluate the reliability of 4D flow MRI of venous and arterial liver hemodynamics for different temporal and spatial resolutions and thus total scan times at 3T and assess repeatability of measurements in healthy volunteers.

Methods

Study Population

Study approval was obtained from our local ethics board and written informed consent was received from all volunteers before MRI investigation. The study population consisted of ten healthy young volunteers without a history of clinical liver disease (8 female and 2 male, mean age=22.3±2.8years).

MR Imaging

All imaging was performed at 3T (Magnetom Trio, Siemens Medical Systems, Erlangen, Germany). Data acquisition consisted of a non-contrast enhanced 3D time-resolved rf-spoiled phase contrast gradient echo sequence with tri-directional velocity encoding (4D flow MRI) (21). Respiratory gating at the lung-spleen interface was applied minimizing ghosting artifacts and image blurring by acquiring an end-diastolic navigator signal for the detection of the lung-spleen boundary at end-expiration. Navigator gating included an acceptance window of 7mm (21). 4D flow MRI used prospective ECG-gating (cine imaging) which combined with k-space segmented data acquisition with Nk=3–4 phase encoding lines for each cardiac time frame. The receive system consisted of a 6-element spine coil and a flexible 6-element chest coil. 4D flow data was acquired in an axial oblique 3D volume angled along the portal vein with complete volumetric coverage of the hepatic arterial and portal venous systems. Velocity encoding sensitivity was set to 100cm/s for all three flow-encoding directions.

Spatio-temporal resolution and total scan time

For the evaluation of different spatial and temporal resolutions, 3 different sequences were examined. The most important differences between sequences 1–3 were as follows: spatial resolution 2.4×2.0×2.4mm3, Nk=3, temporal resolution 61.2ms (sequence 1, longest scan time); spatial resolution 2.5×2.0×2.4mm3, Nk=4, temporal resolution 81.6ms (sequence 2, medium scan time) and spatial resolution 2.6×2.5×2.6mm3, Nk=4, temporal resolution 80ms (sequence 3, shortest scan time). The total scan time and number of cardiac time frames for each sequence depended on the subjects’ heart rate and efficiency of the respiratory navigator gating. A detailed summary of all pulse sequence parameters is provided in Table 1.

Table 1.

Summary of sequence parameters for all 4D flow MRI acquisitions.

Sequence 1 Sequence 2 Sequence 3
Spatial resolution [mm3] 2.4×2.0×2.4 2.5×2.0×2.4 2.6×2.5×2.6
Temporal resolution [ms] 61.2 81.6 80
TE [ms] 2.6 2.6 2.5
Nk 3 4 4
Total scan time [min] 14.6±3.3* 9.8±2.4* 8.2±2.3*
Acquired cardiac time frames/ cycle 12.4±1.8 9.0±1.2* 9.2±1.2*
Velocity encoding sensitivity [cm/s] 100
Flip angle [°] 7
Imaging matrix [mm2] 192×152
Band width [Hz/Pixel] 450
Slices/ slab 36
Parallel imaging (GRAPPA)
  reduction factor R 2
  reference lines 32
Open in a new tab
*

= significant differences were present in scan times between sequences 1, 2 and 3 (p<0.001) and for acquired cardiac time frame/ cycle between sequences 1 vs. 2 and sequences 1 vs. 3 (p<0.001).

Scan-Rescan Reliability

To assess the scan-rescan reproducibility of the MRI measurements all volunteers were scanned twice on the same 3T MRI system with at least 5 months’ time difference between scans. Participants were instructed to fast for 6 hours prior to each scan to match portal physiology at the time of scanning.

In total, each volunteer underwent each 4D flow MRI sequence twice, for a total of n=6 scans.

Data Analysis

To reduce phase offset errors, Maxwell and eddy current corrections were applied prior to 4D flow data analysis. Maxwell correction was performed as described by Bernstein et al (22) during the image reconstruction on the MR system. Next, data pre-processing used a home built analysis tool (programmed in Matlab, the MathWorks, USA) and included velocity anti-aliasing (23), noise filtering and correction for eddy currents as previous described by Walker et al. (24). Briefly, eddy current correction was based on defining a threshold based separation of regions with static tissue from blood flow and noise by using the velocity-time standard deviation for each pixel. Pixels were regarded as static tissue for standard deviations below the adjusted threshold. A 1st order plane fit to the static regions provided an estimation of linearly varying phase offset errors. Afterwards they were subtracted from the images. After correction, a 3D PC MR angiogram (MRA) (25) was calculated to generate an iso-surface rendering (EnSight, CEI, Apex, USA) of the hepatic vascular system. The 3D PC-MRA was used to manually place ten analysis planes in the 3D volume at the following locations: the splenic and superior mesenteric veins, the splenic-mesenteric confluence, the intrahepatic portal vein, the right and left intrahepatic portal vein branches, the celiac trunk, and the splenic, hepatic and superior mesenteric arteries (Figure 1).

Figure 1.

Figure 1

Open in a new tab

3D streamlines visualization of arterial and portal venous maximum velocity-time curves in six analysis planes for sequence 1. The blue 2D analysis planes were manually positioned in the splenic vein, superior mesenteric vein (smv), splenic-mesenteric confluence (spl.-mes. confl.), right (right PV) and left (left PV) intrahepatic portal vein branch, celiac trunk, splenic artery, hepatic artery and superior mesenteric artery (sma). Error bars reflect inter-individual standard deviations among all volunteers (n=10). Color coding = local blood flow velocity.

3D blood flow visualization

The ten analysis planes were used as emitter planes for the calculation of 3D streamlines, which reflect traces along 3-directional blood flow velocities for individual cardiac time-frames. In addition, time-resolved 3D particle traces were generated which showed the temporal dynamics of blood flow over one cardiac cycle (6). 3D display software was used to enable a visual grading of the streamlines and particular traces from any view angle (EnLiten, CEI, Apex, USA). For each of the ten vessels, two independent readers graded streamlines and time-resolved particle traces on a three-point scale (0=not visible, 1=partially visible, 2=completely visible) and noted visualization artifacts such as leakage of the traces into adjacent vessels (present or not present).

3D blood flow quantification

For all 10 analysis planes quantitative hemodynamic analysis was performed by manually delineating the vessel lumen contours for each cardiac time-frame using a home-built tool (programmed in Matlab, The Mathworks, USA). For each plane, maximum and mean velocities, flow volume and vessel lumen area were calculated. All measurements were indirectly evaluated using the concept of conservation of mass to evaluate the internal consistency of the acquired flow data (15). Therefore the blood flow was indirectly measured in joining and branching vessels of the portal venous and arterial system. Blood flow volume in the splenic-mesenteric confluence was calculated by adding the flow volumes from the joining splenic and superior mesenteric veins (flow volume splenic-mesenteric confluence=flow volume splenic vein+flow volume superior mesenteric vein). Blood flow volume in the intrahepatic portal vein was calculated by adding flow from the right and left intrahepatic portal vein branches (flow volume intrahepatic portal vein=flow volume right portal vein branch+flow volume left portal vein branch). In the arterial system we assumed flow volume in the celiac trunk results from the hepatic and splenic arteries (flow volume celiac trunk=flow volume hepatic artery+flow volume splenic artery) neglecting flow from the left gastric artery based on standard anatomy of the volunteers.

Statistical Analysis

Statistical analyses were performed using commercially available software (SPSS 19.0; SPSS, Chicago, Illinois, USA). Continuous variables are presented as mean ± standard deviation. Interobserver agreement was evaluated using Cohen’s kappa statistics. Kappa values were interpreted as follows: 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 substantial agreement, and 0.81–1.00 excellent agreement, as described previously (26).

Continuous variables were evaluated by paired, two tailed t-tests. A p-value <0.05 was considered statistically significant. Scan-rescan reproducibility for all volunteers and analysis planes was assessed by the Bland-Altman approach using the mean difference (d) and standard deviation of the difference (s), calculating the limits of agreement (± 2SD) using 95% confidence intervals (27).

Results

One volunteer did not consent for the rescan evaluation; three 4D flow MRI measurements failed due to technical problems. As a result, a total of n=54 hepatic 4D flow MRI data sets were available for analysis. The heart rate averaged 68.5±6.1bpm across the study cohort and over all examinations. No significant differences in heart rate were seen between scan-rescan measurements or for different spatio-temporal resolution scans.

For each volunteer, total scan times and numbers of acquired cardiac time frames were dependent on heart rate and navigator efficiency. As summarized in Table 1, reduction in spatial-temporal resolution resulted in a significantly shorter average total acquisition time of 33% between sequence 1 and 2 and of 44% between sequence 1 and 3 (p<0.001).

3D visualization of arterial and portal venous hemodynamics

As summarized in Table 2, both readers confirmed successful 3D flow visualization (complete visibility of streamlines and particle traces, grade 2) for the majority of vessels within the splanchnic system. No significant differences for flow visualization were found for different spatio-temporal resolutions or scan versus rescan measurements. Flow visualization was limited for the left intrahepatic branch (incomplete 3D flow visualization in 20 of 54 cases for both readers A and B) and in one case for the hepatic artery (reader A and B) (Table 2). 3D flow visualization artifacts (leakage into adjacent vessels) were frequent and were observed for streamlines in 80% (reader A) and in 69% (reader B), as well as for particle traces in 78% (reader A) and 69% (reader B) of vessels (Table 2). There was excellent agreement between readers, with a Cohen’s kappa of 0.93.

Table 2.

Summary of 3D flow visualization (grade 2 for complete visibility) in 10 volunteers for n=54 4D flow data sets by readers A and B.

visibility
reader A reader B
superior mesenteric vein 100% 100%
splenic vein 100% 100%
splenic-mesenteric confl. 100% 100%
portal vein right branch 100% 100%
portal vein left branch 63% 63%
celiac trunk 100% 100%
hepatic artery 98% 98%
splenic artery 100% 100%
superior mesenteric artery 100% 100%
streamlines leakage present in 80% present in 69%
particle traces leakage present in 78% present in 69%
Open in a new tab

Quantification of arterial and portal venous hemodynamics

Results of flow, velocity and vessel area quantification are summarized in Figure 2 and Table 3. Varied temporal resolution alone (sequence 1 vs. sequence 2) resulted in significant differences in mean velocity for the intrahepatic portal vein (10.8cm/sec vs. 11.3 cm/sec, p=0.013, Figure 2B). Flow volumes were similar for all vessels except for significant differences for the splenic-mesenteric confluence (0.76l/min vs. 0.78l/min, p=0.048) and the hepatic artery (0.36l/min vs. 0.33l/min, p=0.033, Figure 2C).

Figure 2.

Figure 2

Open in a new tab

Flow quantification for all measurements using 4D flow sequences 1–3 for all ten analysis planes. Each bar represents the standard deviation over 10 volunteers and two scan-rescan measurements.

(* = significant difference compared to sequence 1, p<0.05)

(+ = significant difference between sequence 2 and 3, p<0.05)

(smv = superior mesenteric vein, spl.-mes. confl. = splenic-mesenteric confluence, intrahep. PV = intrahepatic portal vein, PV = portal vein, sma = superior mesenteric artery)

Table 3.

Bland and Altman analyses (mean bias ± 2SD) for arterial and portal venous flow parameters in the hepatic and splanchnic system illustrating the agreement between sequence 1 vs. sequence 2 and sequence 1 vs. sequence 3 as well as scan-rescan reproducibility between scan 1 and scan 2.

Bland Altman analysis
[difference in percentage normalized by the mean of the two compared]
peak Velocity seq. 1 vs. seq. 2 seq. 1 vs. seq. 3 scan 1 vs. scan 2
superior mesenteric vein 13.1 ± 24.2 13.2 ± 31.3 −3.4 ± 30.4
splenic vein 3.2 ± 23.7 3.2 ± 26.9 4.1 ± 24.2
splenic-mesenteric confluence −5.4 ± 15.6 7.3 ± 14.5 1.7 ± 26.2
intrahepatic portal vein 6.7 ± 12.8 5.5 ± 13.0 −0.1 ± 20.1
right portal vein branch 9.2 ± 10.6 14.5 ± 13.1 1.3 ± 20.4
left portal vein branch 9.0 ± 16.5 10.1 ± 15.7 0.4 ± 18.2
celiac trunk 10.9 ± 14.3 18.6 ± 12.0 −2.3 ± 24.9
hepatic artery 11.4 ± 15.1 20.6 ± 15.8 5.5 ± 31.8
splenic artery 9.9 ± 24.5 20.1 ± 24.5 0.3 ± 16.9
superior mesenteric artery 2.2 ± 28.3 9.6 ± 27.3 −3.7 ± 40.2
mean Velocity seq. 1 vs. seq. 2 seq. 1 vs. seq. 3 scan 1 vs. scan 2
superior mesenteric vein 8.2 ± 15.8 6.2 ± 14.9 −8.6 ± 34.9
splenic vein 4.5 ± 14.8 0.7 ± 22.6 6.4 ± 22.6
splenic-mesenteric confluence −4.9 ± 11.7 0.4 ± 14.7 1.6 ± 28.5
intrahepatic portal vein 0.4 ± 8.5 4.5 ± 11.1 3.7 ± 22.3
right portal vein branch 0.5 ± 12.8 3.6 ± 11.0 3.5 ± 21.6
left portal vein branch 6.2 ± 16.3 4.8 ± 17.4 5.9 ± 19.7
celiac trunk 0.0 ± 12.9 3.2 ± 16.1 3.4 ± 25.6
hepatic artery 1.0 ± 15.3 8.1 ± 16.8 10.1 ± 27.5
splenic artery 1.8 ± 18.4 6.8 ± 22.6 7.1 ± 24.1
superior mesenteric artery 0.7 ± 14.0 0.9 ± 18.8 −6.4 ± 29.5
flow volume seq. 1 vs. seq. 2 seq. 1 vs. seq. 3 scan 1 vs. scan 2
superior mesenteric vein 7.5 ± 24.8 11.8 ± 37.5 2.2 ± 32.6
splenic vein −6.3 ± 24.9 −5.5 ± 21.9 −2.4 ± 32.0
splenic-mesenteric confluence −1.8 ± 16.7 5.4 ± 22.5 −3.6 ± 23.5
intrahepatic portal vein −0.2 ± 14.5 2.2 ± 19.3 −0.4 ± 21.2
right portal vein branch −0.3 ± 16.3 0.7 ± 20.5 19.3 ± 42.9
left portal vein branch 8.2 ± 15.4 0.5 ± 36.6 −5.6 ± 67.8
celiac trunk 9.7 ± 15.1 16.9 ± 18.7 −9.5 ± 24.3
hepatic artery 16.1 ± 13.9 13.8 ± 10.1 −9.7 ± 36.3
splenic artery 4.4 ± 11.4 14.9 ± 12.2 −7.8 ± 21.5
superior mesenteric artery 16.8 ± 18.8 15.2 ± 22.1 −29.3 ± 35.2
vessel area seq. 1 vs. seq. 2 seq. 1 vs. seq. 3 scan 1 vs. scan 2
superior mesenteric vein 1.4 ± 27.7 3.7 ± 26.3 −0.5 ± 51.0
splenic vein −1.5 ± 35.3 5.4 ± 36.8 −11.7 ± 32.0
splenic-mesenteric confluence 0.3 ± 16.3 5.5 ± 37.2 −8.3 ± 35.1
intrahepatic portal vein 3.8 ± 14.2 −1.0 ± 14.2 −8.6 ± 28.0
right portal vein branch 2.9 ± 27.2 0.2 ± 24.0 8.9 ± 44.8
left portal vein branch −7.9 ± 18.1 −8.0 ± 29.4 −5.7 ± 60.8
celiac trunk 23.1 ± 39.9 25.5 ± 49.0 −14.1 ± 31.5
hepatic artery 6.7 ± 33.8 −6.9 ± 32.6 −17.9 ± 25.4
splenic artery 1.0 ± 31.6 5.1 ± 32.2 −12.0 ± 34.8
superior mesenteric artery −4.6 ± 21.9 −6.4 ± 23.9 −4.3 ± 31.7
Open in a new tab

Decreased spatial and temporal resolution (sequence 1 vs. sequence 3) resulted in significantly different maximum velocities for the right intrahepatic portal vein branch (20.0cm/sec vs. 17.8cm/sec, p=0.007, Figure 2A), the celiac trunk, hepatic and splenic arteries (average mean 19%, range 10%-23%, p<0.05, Figure 2A). Mean velocities were significantly different for the celiac trunk (28.8cm/sec vs. 24.5cm/sec, p=0.009) and hepatic artery (10.9cm/sec vs. 9.4cm/sec, p=0.016, Figure 2B). Flow volumes were similar except for the hepatic artery (0.36l/min vs. 0.32l/min, p=0.007) and splenic artery (0.42l/min vs. 0.38l/min, p=0.027, Figure 2C).

Bland-Altman analysis demonstrated increasing average mean differences in the hepatic arterial system for sequence 1 vs. sequence 2 compared to sequence 1 vs. sequence 3 for the maximum velocities (8% vs. 12%) and mean velocities (3% vs. 4%). Similar average mean differences were shown for flow volumes in the portal venous system (4% vs. 3%) and arterial system (12% vs. 15%) for both comparisons (Table 3).

Scan-rescan analyses across all sequences revealed average differences for the maximum velocities of 2% for the portal venous system and 3% for the arterial system. For mean velocities, average differences of 5% in the portal venous system and 7% in the arterial system were found. Flow volume evaluation revealed average differences of 6% for the portal venous system and 14% for the arterial system mean flow values (Table 3).

Evaluation of the internal consistency of the 4D flow MRI data based on the concept of conservation of mass showed for the flow volume in the splenic-mesenteric confluence an error of 8.8±16.2% (BA=-0.077±0.249l/min). For the intrahepatic portal vein, an error of 3.1±15.4% was found (BA=-0.041±0.241l/min). In the arterial system an error of 1.3±17.5% was shown (BA=-0.004±0.231l/min) for calculation of the celiac axis flow.

Discussion and Conclusions

4D flow MRI of the liver hemodynamics demonstrated good image quality with a low interobserver variability. Varying spatio-temporal resolutions resulted in significantly lower total scan times but also significantly altered velocities and flow volumes particularly in the arterial system. Quantitative assessment of flow parameters offered good scan-rescan reproducibility at 3T except for flow volume data.

Assessment of hepatic blood flow via the portal venous system, a steady flow system with relatively low blood flow and low flow velocities, has gained growing relevance for patients with portal hypertension for assessment of disease severity and planning therapies (1, 28, 29). Doppler US represents the clinical reference standard for therapeutic control of portal hypertension with known limitations, principle among which is interobserver variability (30,31). 2D cine PC MRI is an alternative technique for hemodynamic assessment of the liver vasculature with low intra- and interobserver variability and higher reproducibility compared to Doppler US (5,29,32). The technique, however, is limited to local 2D acquisition planes. Flow measurements and potential flow quantification obtained from 2D cine PC MRI can be inaccurate due to difficulties with precise plane positioning in the complex liver vasculature (33).

In contrast, 4D flow MRI offers the possibility to retrospectively position flow analysis planes throughout the entire 3D volume while enabling comprehensive evaluation of liver hemodynamics as visualized by 3D streamlines and time-resolved 3D particle traces. For 3D flow visualization in our study cohort, all portal vessels were well visualized except the left intrahepatic portal vein branch, likely a consequence of the vessel size (Table 2). However, 3D visualization artifacts (leakage of traces into adjacent vessels) were frequent and were seen up to 80% of cases at all spatial and temporal resolutions. We hypothesize that these visualization limitations are based on spatio-temporal blurring as a result of respiratory gating during data acquisition with free breathing and thus reduced effective resolution for all three sequences. Given the low pulsatile portal flow pattern, the nominal spatial resolution of the imaging protocols should have been sufficient for the visualization of flow in the right and left intrahepatic portal vein branches. Future studies should thus include a more detailed investigation of the impact of further improvements in spatial resolution for 3D visualization quality.

Quantitative assessment of all data using different scan protocols with lower spatio-temporal resolution resulted in significantly lower maximum and mean velocities, especially in the arterial system. The lower velocity portal and splanchnic venous systems showed significant differences in maximum velocity for the right intrahepatic vein branch and in mean velocity for the intrahepatic portal vein. Thus, lower spatio-temporal resolution cannot be used for hemodynamic analysis for the arterial and PV system and scan time savings cannot be achieved.In order to achieve reproducible quantitative measurements in both the arterial and venous vessels in a reasonable amount of time more advanced acceleration techniques are necessary to accelerate the 4D flow MRI acquisition. Possible solutions include compressed sensing (34) or k-t under sampling (35) along spatial and temporal dimensions yielding a reduced scan time or a higher spatial resolution while keeping the total scan time constant (36).

Looking at the accuracy of hemodynamic data acquired with 4D flow MRI a recently reported reproducibility study of 4D flow MRI in the thoracic aorta with a high blood flow velocities revealed a good scan-rescan reliability as well as inter- and intra-observer comparison for clinically relevant blood flow parameters and wall shear stress (19). Frydrychowicz et al. qualitatively visualized hepatic blood flow using a 3D radial under sampling technique with high spatial resolution and good volumetric coverage (13). A follow-up study revealed a good internal consistency and inter- and intraobserver variability for quantifying hepatic and splanchnic hemodynamics in patients with portal hypertension (15). However, in this study the accuracy of the flow measurements was evaluated indirectly by applying internal consistency and conservation of mass principles; no repeatability experiments were performed (15). In our study, indirect evaluation of the internal consistency of the quantitative 4D flow MRI analysis using the conservation of mass principle revealed relatively small errors. This indicates internal consistency of flow volume measurements in the splenic-mesenteric confluence, intrahepatic portal vein and celiac trunk.

Scan-rescan repeatability validation performed at least 5 months apart revealed good reproducibility for 4D flow MRI quantification in the portal venous system with lower blood flow velocities. Quantification of arterial maximum velocities by 4D flow MRI showed robust results, but reproducibility for flow volume measurements was limited. In the literature Doppler US for the liver vasculature, representing the current clinical standard, shows an interobserver variability between 16% and 26% for the portal blood flow velocity (37). In our study we found a similar value for scan-rescan variability of 25% and 26% for maximum and mean velocities. A possible reason for these variations might be related to the manual segmentation of vessel lumen boundaries. Semi- or fully automated segmentation techniques may offer the possibility to improve the reproducibility of the calculation of parameters depending on the vessel area (38).

A further strength of 4D flow MRI measurements compared to Doppler Ultrasound is the ability to quantify flow in a robust manner. It is also possible to perform internal consistency checks which are crucial to the understanding of the data quality in the individual patient.

This study has several limitations. The principle limitation is the lack of a gold standard measurement for comparison of hemodynamic data. The rationale to not pursue US correlation was based on limited inter-observer reproducibility of the US Doppler-based hemodynamic measurements reported in the literature. Furthermore the rationale to not use a 2D PC MRI correlation was due to its difficulties with precise plane positioning in the complex liver vasculature confounding utility as a gold standard. Finally, our study cohort was comprised exclusively of healthy volunteers, limiting direct translation of our results in a patient cohort.

In conclusion, 4D flow MRI enables a comprehensive assessment of liver hemodynamics in both the hepatic arterial and portal venous vasculature with similar inter-observer variability compared to Doppler ultrasound. The strength of 4D flow MRI compared to Doppler US is the ability to quantify flow in a robust manner. The volumetric acquisition permits retrospective interrogation of all components of the hepatic vascular bed. Quantitative flow parameters showed good scan-rescan reproducibility only limited by the measurements of flow volume data. 4D flow acquisition with lower spatio-temporal resolution cannot be used for hemodynamic analyses for the arterial and PV system and, thus, scan time savings have to be achieved differently. Future studies should aim to apply 4D flow MRI in a clinical setting with reduced scan time and improved spatio-temporal resolution, with the application of advanced acceleration techniques. Further investigations are necessary to validate the accuracy of the 4D flow MRI data using an independent flow quantification method, taking into consideration the advantages and disadvantages of each when determining quantitative accuracy. Investigations based on patients with increased variability of liver hemodynamics are warranted for further evaluation of the scan-rescan reliability in a clinical setting and correlation of hemodynamic data with parameters of portal hypertension and liver cirrhosis for monitoring disease progression and assessment of mean therapy effects.

Acknowledgements

Grant Support: Supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number R01HL115828 and the Northwestern Memorial Foundation Dixon Translational Research Grants Initiative. Supported by a DFG fellowship of the German Research Foundation (DFG) under Award Number STA 1288/ 2-1.

Literature

  • 1.Groszmann RJ. Hyperdynamic circulation of liver disease 40 years later: pathophysiology and clinical consequences. Hepatology. 1994;20:1359–1363. [PubMed] [Google Scholar]
  • 2.Abraldes JG, Tarantino I, Turnes J, Garcia-Pagan JC, Rodés J, Bosch J. Hemodynamic response to pharmacological treatment of portal hypertension and long-term prognosis of cirrhosis. Hepatology. 2003;37:902–908. doi: 10.1053/jhep.2003.50133. [DOI] [PubMed] [Google Scholar]
  • 3.Garcia-Tsao G, Bosch J. Management of varices and variceal hemorrhage in cirrhosis. N Engl J Med. 2010;362:823–832. doi: 10.1056/NEJMra0901512. [DOI] [PubMed] [Google Scholar]
  • 4.Wigstrom L, Sjoqvist L, Wranne B. Temporally resolved 3D phase-contrast imaging. Magn Reson Med. 1996;36:800–803. doi: 10.1002/mrm.1910360521. [DOI] [PubMed] [Google Scholar]
  • 5.Gouya H, Vignaux O, Sogni P, et al. Chronic liver disease: systemic and splanchnic venous flow mapping with optimized cine phase-contrast MR imaging validated in a phantom model and prospectively evaluated in patients. Radiology. 2011;261:144–155. doi: 10.1148/radiol.11101541. [DOI] [PubMed] [Google Scholar]
  • 6.Buonocore MH. Visualizing blood flow patterns using streamlines, arrows, and particle paths. Magn Reson Med. 1998;40:210–226. doi: 10.1002/mrm.1910400207. [DOI] [PubMed] [Google Scholar]
  • 7.Kozerke S, Hasenkam JM, Pedersen EM, Boesiger P. Visualization of flow patterns distal to aortic valve prostheses in humans using a fast approach for cine 3D velocity mapping. J Magn Reson Imaging. 2001;13:690–698. doi: 10.1002/jmri.1097. [DOI] [PubMed] [Google Scholar]
  • 8.Sanz J, Kuschnir P, Rius T, et al. Pulmonary arterial hypertension: noninvasive detection with phase-contrast MR imaging. Radiology. 2007;243:70–79. doi: 10.1148/radiol.2431060477. [DOI] [PubMed] [Google Scholar]
  • 9.Ley S, Unterhinninghofen R, Ley-Zaporozhan J, Schenk JP, Kauczor HU, Szabo G. Validation of magnetic resonance phasecontrast flow measurements in the main pulmonary artery and aorta using perivascular ultrasound in a large animal model. Invest Radiol. 2008;43:421–426. doi: 10.1097/RLI.0b013e318169015d. [DOI] [PubMed] [Google Scholar]
  • 10.Harloff A, Albrecht F, Spreer J, et al. 3D blood flow characteristics in the carotid artery bifurcation assessed by flow-sensitive 4D MRI at 3T. Magn Reson Med. 2009;61:65–74. doi: 10.1002/mrm.21774. [DOI] [PubMed] [Google Scholar]
  • 11.Markl M, Frydrychowicz A, Kozerke S, Hope M, Wieben O. 4D flow MRI. J Magn Reson Imaging. 2012;36:1015–1036. doi: 10.1002/jmri.23632. [DOI] [PubMed] [Google Scholar]
  • 12.Stankovic Z, Frydrychowicz A, Csatari Z, et al. MR-based visualization and quantification of three-dimensional flow characteristics in the portal venous system. J Magn Reson Imaging. 2010;32:466–475. doi: 10.1002/jmri.22248. [DOI] [PubMed] [Google Scholar]
  • 13.Frydrychowicz A, Landgraf BR, Niespodzany E, et al. Four-Dimensional Velocity Mapping of the Hepatic and Splanchnic Vasculature With Radial Sampling at 3 Tesla: A Feasibility Study in Portal Hypertension. J Magn Reson Imaging. 2011;34:577–584. doi: 10.1002/jmri.22712. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Stankovic Z, Csatari Z, Deibert P, et al. Normal and Altered 3D Portal Venous Hemodynamics in Patients with Liver Cirrhosis. Radiology. 2012;262:862–873. doi: 10.1148/radiol.11110127. [DOI] [PubMed] [Google Scholar]
  • 15.Roldán-Alzate A, Frydrychowicz A, Niespodzany E, et al. In vivo validation of 4D flow MRI for assessing the hemodynamics of portal hypertension. J Magn Reson Imaging. 2013;37:1100–1108. doi: 10.1002/jmri.23906. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Stankovic Z, Csatari Z, Deibert P, et al. A feasibility study to evaluate splanchnic arterial and venous hemodynamics by flow-sensitive 4D MRI compared with Doppler ultrasound in patients with cirrhosis and controls. Eur J Gastroenterol Hepatol. 2013;25:669–675. doi: 10.1097/MEG.0b013e32835e1297. [DOI] [PubMed] [Google Scholar]
  • 17.Uribe S, Beerbaum P, Sorensen TS, Rasmusson A, Razavi R, Schaeffter T. Four-dimensional (4D) flow of the whole heart and great vessels using real-time respiratory self-gating. Magn Reson Med. 2009;62:984–992. doi: 10.1002/mrm.22090. [DOI] [PubMed] [Google Scholar]
  • 18.Stalder AF, Russe MF, Frydrychowicz A, Bock J, Hennig J, Markl M. Quantitative 2D and 3D phase contrast MRI: optimized analysis of blood flow and vessel wall parameters. Magn Reson Med. 2008;60:1218–1231. doi: 10.1002/mrm.21778. [DOI] [PubMed] [Google Scholar]
  • 19.Markl M, Wallis W, Harloff A. Reproducibility of Flow and Wall Shear Stress Analysis Using Flow-Sensitive Four-Dimensional MRI. J Magn Reson Imaging. 2011;33:988–994. doi: 10.1002/jmri.22519. [DOI] [PubMed] [Google Scholar]
  • 20.Strecker C, Harloff A, Wallis W, Markl M. Flow-sensitive 4D MRI of the thoracic aorta: Comparison of image quality, quantitative flow, and wall parameters at 1.5 T and 3T. J Magn Reson Imaging. 2012;36:1097–1103. doi: 10.1002/jmri.23735. [DOI] [PubMed] [Google Scholar]
  • 21.Markl M, Harloff A, Bley TA, et al. Time-resolved 3D MR velocity mapping at 3T: improved navigator-gated assessment of vascular anatomy and blood flow. J Magn Reson Imaging. 2007;25:824–831. doi: 10.1002/jmri.20871. [DOI] [PubMed] [Google Scholar]
  • 22.Bernstein MA, Zhou XJ, Polzin JA, et al. Concomitant gradient terms in phase contrast MR: analysis and correction. Magn Reson Med. 1998;39:300–308. doi: 10.1002/mrm.1910390218. [DOI] [PubMed] [Google Scholar]
  • 23.Bock J, Kreher BW, Hennig J, Markl M. Optimized pre-processing of time-resolved 2D and 3D Phase Contrast MRI data; Paper presented at: 15th International Society For Magnetic Resonance In Medicine 2007, paper number 3138. [Google Scholar]
  • 24.Walker PG, Cranney GB, Scheidegger MB, Waseleski G, Pohost GM, Yoganathan AP. Semiautomated method for noise reduction and background phase error correction in MR phase velocity data. J Magn Reson Imaging. 1993;3:521–530. doi: 10.1002/jmri.1880030315. [DOI] [PubMed] [Google Scholar]
  • 25.Bock J, Frydrychowicz A, Stalder AF, et al. 4D phase contrast MRI at 3 T: effect of standard and blood-pool contrast agents on SNR, PC-MRA, and blood flow visualization. Magn Reson Med. 2010;63:330–338. doi: 10.1002/mrm.22199. [DOI] [PubMed] [Google Scholar]
  • 26.Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159–174. [PubMed] [Google Scholar]
  • 27.Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–310. [PubMed] [Google Scholar]
  • 28.Zoli M, Marchesini G, Cordiani MR, et al. Echo-Doppler measurement of splanchnic blood flow in control and cirrhotic subjects. J Clin Ultrasound. 1986;14:429–435. doi: 10.1002/jcu.1870140605. [DOI] [PubMed] [Google Scholar]
  • 29.Burkart DJ, Johnson CD, Morton MJ, Wolf RL, Ehman RL. Volumetric flow rates in the portal venous system: measurement with cine phase-contrast MR imaging. AJR Am J Roentgenol. 1993;160:1113–1118. doi: 10.2214/ajr.160.5.8470589. [DOI] [PubMed] [Google Scholar]
  • 30.Burns PN, Jaffe CC. Quantitative flow measurements with Doppler ultrasound: techniques, accuracy, and limitations. Radiol Clin North Am. 1985;23:641–657. [PubMed] [Google Scholar]
  • 31.de Vries PJ, van Hattum J, Hoekstra JB, de Hooge P. Duplex Doppler measurements of portal venous flow in normal subjects. Inter- and intra-observer variability. J Hepatol. 1991;13:358–363. doi: 10.1016/0168-8278(91)90081-l. [DOI] [PubMed] [Google Scholar]
  • 32.Yzet T, Bouzerar R, Allart JD, et al. Hepatic vascular flow measurements by phase contrast MRI and doppler echography: a comparative and reproducibility study. J Magn Reson Imaging. 2010;31:579–588. doi: 10.1002/jmri.22079. [DOI] [PubMed] [Google Scholar]
  • 33.Buonocore MH, Bogren H. Factors influencing the accuracy and precision of velocity-encoded phase imaging. Magn Reson Med. 1992;26:141–154. doi: 10.1002/mrm.1910260115. [DOI] [PubMed] [Google Scholar]
  • 34.Tariq U, Hsiao A, Alley M, Zhang T, Lustig M, Vasanawala SS. Venous and arterial flow quantification are equally accurate and precise with parallel imaging compressed sensing 4D phase contrast MRI. J Magn Reson Imaging. 2013 Jun;37(6):1419–1426. doi: 10.1002/jmri.23936. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Jung B, Stalder AF, Bauer S, Markl M. On the undersampling strategies to accelerate time-resolved 3D imaging using k-t-GRAPPA. Magn Reson Med. 2011;66:966–975. doi: 10.1002/mrm.22875. [DOI] [PubMed] [Google Scholar]
  • 36.Baltes C, Kozerke S, Hansen MS, Pruessmann KP, Tsao J, Boesiger P. Accelerating cine phase-contrast flow measurements using k-t BLAST and k-t SENSE. Magn Reson Med. 2005;54:1430–1438. doi: 10.1002/mrm.20730. [DOI] [PubMed] [Google Scholar]
  • 37.Sabbà C, Merkel C, Zoli M, et al. Interobserver and interequipment variability of echo-Doppler examination of the portal vein: effect of a cooperative training program. Hepatology. 1995;21:428–433. doi: 10.1002/hep.1840210225. [DOI] [PubMed] [Google Scholar]
  • 38.van der Geest RJ, Niezen RA, van der Wall EE, de Roos A, Reiber JH. Automated measurement of volume flow in the ascending aorta using MR velocity maps: evaluation of inter- and intraobserver variability in healthy volunteers. J Comput Assist Tomogr. 1998;22:904–911. doi: 10.1097/00004728-199811000-00013. [DOI] [PubMed] [Google Scholar]

RESOURCES