Eur Radiol DOI 10.1007/s00330-015-3663-x
VASCULAR-INTERVENTIONAL
Effect of TIPS placement on portal and splanchnic arterial blood flow in 4-dimensional flow MRI Zoran Stankovic & Martin Rössle & Wulf Euringer & Michael Schultheiss & Riad Salem & Alex Barker & James Carr & Mathias Langer & Michael Markl & Jeremy D. Collins
Received: 14 July 2014 / Revised: 7 February 2015 / Accepted: 11 February 2015 # European Society of Radiology 2015
Abstract Objectives To assess changes in portal and splanchnic arterial haemodynamics in patients undergoing transjugular intrahepatic portosystemic shunt (TIPS) using fourdimensional (4D) flow MRI, a non-invasive, non-contrast imaging technique. Methods Eleven patients undergoing TIPS implantation were enrolled. K-t GRAPPA accelerated non-contrast 4D flow MRI of the liver vasculature was applied with acceleration factor R= 5 at 3Tesla. Flow analysis included three-dimensional (3D) blood flow visualization using time-resolved 3D particle traces and semi-quantitative flow pattern grading. Quantitative evaluation entailed peak velocities and net flows throughout the arterial and portal venous (PV) systems. MRI measurements were taken within 24 h before and 4 weeks after TIPS placement. Results Three-dimensional flow visualization with 4D flow MRI revealed good image quality with minor limitations in PV flow. Quantitative analysis revealed a significant increase in PV flow (562±373 ml/min before vs. 1831±965 ml/min
after TIPS), in the hepatic artery (176±132 ml/min vs. 354± 140 ml/min) and combined flow in splenic and superior mesenteric arteries (770 ml/min vs. 1064 ml/min). Shunt-flow assessment demonstrated stenoses in two patients confirmed and treated at TIPS revision. Conclusions Four-dimensional flow MRI might have the potential to give new information about the effect of TIPS placement on hepatic perfusion. It may explain some unexpected findings in clinical observation studies. Key Points • 4D flow MRI, a non-invasive, non-contrast imaging technique, is feasible after TIPS. • Provides visualization and quantification of hepatic arterial, portal venous, collateral and TIPS haemodynamics. • Better understanding of liver blood flow changes after TIPS and patient management.
Z. Stankovic : R. Salem : A. Barker : J. Carr : M. Markl : J. D. Collins Department of Radiology, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA
Abbreviations NO ET-1 TIPS Doppler US MRI 4D flow MRI
Z. Stankovic (*) : W. Euringer : M. Langer Department of Diagnostic Radiology and Medical Physics, University Medical Center Freiburg, Hugstetter Strasse 55, 79106 Freiburg, Germany e-mail: [email protected]
Keywords 4D flow MRI . Liver cirrhosis . Liver haemodynamics . TIPS . Splanchnic system
M. Rössle : M. Schultheiss Department of Gastroenterology, University Medical Center Freiburg, Freiburg, Germany
IVC MELD GRAPPA
M. Markl Department of Biomedical Engineering, McCormick School of Engineering, Northwestern University, Chicago, IL, USA
PC-MRA HVPG
Nitric oxide Endothelin-1 Transjugular intrahepatic portosystemic shunt Doppler Ultrasound Magnetic resonance imaging Time-resolved (CINE) phase contrast (PC) gradient echo sequence with 3-directional and 3-dimensional velocity mapping Inferior vena cava Model for End-Stage Liver Disease Generalized auto-calibrating partially parallel acquisitions 3D phase-contrast MR angiogram Hepatic venous pressure gradient
Eur Radiol
Introduction Symptomatic portal hypertension leads to several clinical symptoms such as variceal bleeding and ascites which often occurs when portosystemic pressure gradients exceed 12 mmHg [1]. It is commonly accepted that increases in portal pressure result from structural and dynamic components. Architectural changes in the liver cause an increase in vascular resistance, the morphological basis of the ‘backward’ component of portal hypertension [2, 3]. It may be augmented by dynamic and reversible constriction of sinusoids by hepatic stellate cells, postulated to be secondary to faulty metabolism of nitric oxide (NO) and imbalanced vasoactive compounds such as endothelin-1 (ET-1), angiotensin II, catecholamines and leukotrienes [4, 5]. A forward flow component has also been postulated. Vasodilating agents such as NO, produced by the congested gut and under the influence of bacterial toxins, escape hepatic clearance and lead to the systemic vasodilatation most evident in the splanchnic blood vessels. This raises the splanchnic blood flow to a degree that augments portal hypertension significantly [6–9]. According to this theory, sinusoidal (structural) hypertension initiates portal hypertension but is itself not the dominant factor. The forward flow theory is based on numerous studies in animal models of various models of cirrhosis, portal vein obstruction and chronic schistosomiasis [5, 10]. Changes in the splanchnic and portal systems are accompanied by a decrease in peripheral resistance and an increase in plasma volume and cardiac output. Most of the clinical haemodynamic alterations in patients with cirrhosis can be explained by simply translating the animal findings to humans. However, reliable measurements in splanchnic arterial flows in humans have not been taken to date because of methodological difficulties in quantifying the flow in splanchnic arterial vessels. To further investigate the haemodynamic alterations in the splanchnic system in patients with chronic portal hypertension, simultaneous measurement of portal and arterial flows may help us prove the forward flow hypothesis and quantify collateral blood flow as the difference between arterial and portal flows in the individual patient. Duplex sonography may not enable us to provide reliable data because measurements of the flow in splanchnic arteries are compromised by the difficulty of measuring vessel diameters. Moreover, duplex sonography strongly depends on the investigator’s skill and is hampered by conditions such as ascites, excessive body weight and bowel gas, all of which restrict its efficacy during quantitative splanchnic haemodynamic assessment. A potential alternative diagnostic imaging approach based on non-contrast four-dimensional (4D) flow MRI enables us to visualize and quantify the blood flow in various regions of the portal and splanchnic arterial systems [11–17], including transjugular intrahepatic portosystemic shunts (TIPS) [18]. The method may also be used to calculate pressure gradients [19, 20]. Thus, 4D flow MRI may allow ‘one-stop-shopping’
to obtain the entire spectrum of information on portal, arterial, collateral and shunt haemodynamics. The purpose of this study was to quantify hepatic arterial and portal haemodynamics in patients with liver cirrhosis using noncontrast 4D flow MRI at 3 T and to investigate changes in hepatic haemodynamics following TIPS insertion.
Materials and methods Study population Written informed consent for participation in this prospective study was obtained from each patient before MRI examination, and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by our local ethics boards. Eleven patients with advanced cirrhosis were included, eight at the University Medical Center, Freiburg, Germany and three at the Northwestern University in Chicago, IL, USA. The study cohort underwent a total of 22 4D flow MRI examinations. All 11 patients were examined before and 4 weeks after TIPS implantation. The aetiologies of cirrhosis were alcohol abuse (n=9), chronic viral hepatitis (n=1) and primary biliary cirrhosis (n=1). Covered stents where implanted preferentially. However, in patients with a high risk of shunt-induced side effects, e.g. older patients, patients with previous episodes of hepatic encephalopathy, patients with bilirubin concentrations of greater than 3 mg/ dl or patients with a Child-Pugh score of >9, bare stents were implanted. As a result four patients received a bare stent (Nitinol Stent System OTW, Eucatech AG, Rheinfelden, Germany; Sinus-SuperFlex-Visual 6 F, Optimed, Ettlingen, Germany), while seven patients received a covered TIPS stent (Viatorr, W. L. Gore & Associates Inc., Flagstaff, AZ, USA). In most of the cases (nine out of 11) the reason for the TIPS placement was refractory ascites, in two cases the reason was variceal bleeding. Ten patients (94 %) received a stent of 10-mm diameter while one obtained a 12-mm stent. Stent length was 6 cm in nine cases (82 %) and 4 cm and 7 cm in one patient each. All measurements were taken after overnight fasting. The patients’ biometrical and biochemical characteristics are shown in Table 1. MR imaging A non-contrast-enhanced three-dimensional (3D) time-resolved (cine) rf-spoiled phase contrast (PC) gradient echo sequence with three-directional velocity encoding (4D flow MRI) was used to assess liver haemodynamics [21, 22]. 4D flow MRI was performed on a 3 T MR system (Magnetom TRIO and Magnetom SKYRA, Siemens, Erlangen, Germany) using a flexible six-element chest coil and six-element spine coil representing the receiver system. All scans were taken during free breathing
Eur Radiol Table 1
Characteristics of the patients before TIPS-implantation Patients (n=11)
Age (years) Sex (female/ male) Body weight (kg) Height (cm) Heart rate (bpm) Blood pressure before MRI (mmhg) Blood pressure after MRI (mmhg) Bilirubin (mg/dl) Creatinine (mg/dl) Albumin (g/dl) INR Refractory ascites (n) Clinical hepatic encephalopathy (n) Child-Pugh class A/B/C (n) Child-Pugh score MELD score
62.5±9.0 3/8 71.6±11.0 171.6±10.8 65.3±6.2 113/68±9/7 113/69±9/6 2.0±1.2 1.0±0.3 3.0±0.4 1.3±0.3 10 0 2/8/1 7.7±1.9 10.9±4.1
with navigator gating at the lung-spleen interface. An enddiastolic navigator signal detected the lung-spleen boundary, which was used for respiratory gating to end-expiration (acceptance window=7 mm) in order to minimize image blurring and ghosting artefacts. 4D flow MRI used prospective ECG-gating combined with k-space segmented data acquisition with Nk =4 phase encoding lines for each cardiac time frame for synchronizing data acquisition during the cardiac cycle. Velocity encoding was obtained by successively executing one reference and three velocity sensitive scans with a temporal resolution of 80 ms and a spatial resolution of 2.4×2.1×2.6 mm3 (TE=2.6 msec, TR= 5 ms). Four-D flow MRI data were acquired in an axial oblique 3D volume angulated along the portal vein with complete volumetric coverage of the hepatic arterial and portal venous systems including the TIPS and outflow in the hepatic veins and inferior vena cava (IVC). Velocity-encoding sensitivity was optimized for the high flow in the hepatic arterial system and TIPS with a value of 100 cm/s in all three flowencoding directions. To shorten the imaging time, we took all scans with k-t GRAPPA acceleration factor R=5. Further parameters were: FOV=260×320 mm2, imaging matrix= 210×160, flip angle α=7°, band width=450 Hz/pixel. The number of cardiac time frames and total imaging time for each examination depended on the efficacy of the respiratory navigator gating and subjects’ heart rates. Data analysis To reduce phase offset errors, Maxwell and eddy current corrections were applied prior to 4D flow data analysis. Maxwell correction was accomplished as described by Bernstein et al. [23] during image reconstruction on the MR system. Data was
subsequently pre-processed using a home-built analysis tool (Matlab, the MathWorks, Natick, MA, USA). This involved velocity anti-aliasing [24], noise filtering and correction for eddy currents as described by Walker et al. [25]. After preprocessing, a 3D phase-contrast (PC) MR angiogram (MRA) was derived from the 4D flow MRI data as described previously [26]. To obtain 3D segmentation of the arterial, portal venous and venous systems, the 3D PC-MRA data was imported to commercial segmentation software (Mimics, Materialise NV, Plymouth, MI, USA) to isolate the velocity data in each vascular region (Fig. 1a). The consequent velocity masks of the 3D PC-MRA were used to depict hepatic arterial and portal venous vascular geometry using iso-surface rendering (EnSight, CEI, Apex, NC, USA). 3D PC-MRA was used to manually place ten analysis planes in the arterial and portal venous systems at the following locations: the splenic and superior mesenteric veins, splenic-mesenteric confluence, proximal portal vein, right and left intrahepatic portal vein branches, coeliac trunk, and splenic, hepatic and superior mesenteric arteries (Fig. 1b). 3D blood flow visualization For 3D flow visualization, all ten analysis planes were used as emitter planes to generate time-resolved 3D particle traces that displayed the temporal dynamics of blood flow over one cardiac cycle [27]. All datasets were evaluated using 3D viewing software (EnLiten, CEI, Apex, NC, USA) enabling visual flow pattern grading by interactive rotation and inspection of the time-resolved particle traces from any view angle. For each of the ten vessels, two independent, experienced readers (Z.S., radiologist, nine years’ abdominal MRI experience; J.C., radiologist, nine years’ cardiovascular MRI experience) semi-quantitatively evaluated image quality of the PC-MRAs and particle traces from the 4D flow MRI datasets. The PCMRA quality grading was based on the vessels’ overall visibility in the portal venous and arterial systems of the liver in all 11 patients before and after TIPS according to a three-point Likert scale (0=not visible, 1=partially visible, 2=completely visible) [28]. Readers were blinded to each other's results, patientidentifying information and the severity of liver disease. 3D blood flow quantification We conducted retrospective quantitative haemodynamic parameter analyses on seven analysis planes (arterial: coeliac trunk, hepatic artery, splenic artery and superior mesenteric artery; portal venous: splenic vein, superior mesenteric vein and portal vein). Net flow over the cardiac cycle and peak velocities were calculated for all 11 patients (baseline and 1 month follow-up examination). The data was acquired during the R-R interval and averaged over multiple cardiac cycles. The 4D flow MRI measurements represent 80–90 % of
Eur Radiol
Fig. 1 4D flow MRI in a 60-year-old male with liver cirrhosis 1 month after TIPS placement. a 3D segmentation of the splanchnic and liver haemodynamics (red: arterial system; blue: venous system; yellow: portal venous system; green: TIPS). abd. aorta=abdominal aorta; spl. artery=splenic artery; spl. vein=splenic vein, hep. artery=hepatic artery;
sma=superior mesenteric artery; smv=superior mesenteric vein; IVC=inferior vena cava. b Colour-coded 3D particle-traces visualization demonstrates increased velocities within the portal vein and TIPS. Ten blue analysis planes were positioned throughout the arterial and portal venous systems to quantify liver haemodynamics
the blood flow within the cardiac cycle including all of the systole and the majority of the diastole. For the constant flow of the portal venous system, corrections are performed based on the acquired time frames extrapolating the constant flow to cover 100 % of the R-R interval. Using 3D segmentation, a home-built analysis tool was applied (Matlab, the MathWorks, Natick, MA, USA) to calculate the peak and mean velocities within the stent. The entire stent was evaluated including the inflow in the stent’s leading end at the portal venous branch through its trailing end within the hepatic vein. In line with the Doppler ultrasound (US) literature, we applied peak shunt velocities below 50 cm/s as cutoff to separate patients with and without shunt stenosis [29, 30]. In addition, we performed invasive hepatic venous pressure gradient (HVPG) measurements before and after TIPS stent graft placement. HVPG represents the gradient between pressure measurements in the intra-abdominal part of inferior vena cava and the portal vein. During the intervention the catheter was inserted into the right jugular vein, passing the right atrium, into the inferior vena cava and into the portal vein.
Results
Statistical analysis Statistical analysis was performed using commercially available software (SPSS 21.0; SPSS, Chicago, IL, USA). Continuous variables are presented as mean±standard deviation. Inter-observer agreement was assessed using Cohen’s kappa statistics with kappa values below 0.20 representing poor agreement, 0.21–0.40 fair agreement, 0.41–0.60 modest agreement, 0.61–0.80 good agreement and 0.81–1.00 excellent agreement [31]. Continuous variables were evaluated by paired, two-tailed t-tests. A p-value
VASCULAR-INTERVENTIONAL
Effect of TIPS placement on portal and splanchnic arterial blood flow in 4-dimensional flow MRI Zoran Stankovic & Martin Rössle & Wulf Euringer & Michael Schultheiss & Riad Salem & Alex Barker & James Carr & Mathias Langer & Michael Markl & Jeremy D. Collins
Received: 14 July 2014 / Revised: 7 February 2015 / Accepted: 11 February 2015 # European Society of Radiology 2015
Abstract Objectives To assess changes in portal and splanchnic arterial haemodynamics in patients undergoing transjugular intrahepatic portosystemic shunt (TIPS) using fourdimensional (4D) flow MRI, a non-invasive, non-contrast imaging technique. Methods Eleven patients undergoing TIPS implantation were enrolled. K-t GRAPPA accelerated non-contrast 4D flow MRI of the liver vasculature was applied with acceleration factor R= 5 at 3Tesla. Flow analysis included three-dimensional (3D) blood flow visualization using time-resolved 3D particle traces and semi-quantitative flow pattern grading. Quantitative evaluation entailed peak velocities and net flows throughout the arterial and portal venous (PV) systems. MRI measurements were taken within 24 h before and 4 weeks after TIPS placement. Results Three-dimensional flow visualization with 4D flow MRI revealed good image quality with minor limitations in PV flow. Quantitative analysis revealed a significant increase in PV flow (562±373 ml/min before vs. 1831±965 ml/min
after TIPS), in the hepatic artery (176±132 ml/min vs. 354± 140 ml/min) and combined flow in splenic and superior mesenteric arteries (770 ml/min vs. 1064 ml/min). Shunt-flow assessment demonstrated stenoses in two patients confirmed and treated at TIPS revision. Conclusions Four-dimensional flow MRI might have the potential to give new information about the effect of TIPS placement on hepatic perfusion. It may explain some unexpected findings in clinical observation studies. Key Points • 4D flow MRI, a non-invasive, non-contrast imaging technique, is feasible after TIPS. • Provides visualization and quantification of hepatic arterial, portal venous, collateral and TIPS haemodynamics. • Better understanding of liver blood flow changes after TIPS and patient management.
Z. Stankovic : R. Salem : A. Barker : J. Carr : M. Markl : J. D. Collins Department of Radiology, Feinberg School of Medicine, Northwestern University, Chicago, IL, USA
Abbreviations NO ET-1 TIPS Doppler US MRI 4D flow MRI
Z. Stankovic (*) : W. Euringer : M. Langer Department of Diagnostic Radiology and Medical Physics, University Medical Center Freiburg, Hugstetter Strasse 55, 79106 Freiburg, Germany e-mail: [email protected]
Keywords 4D flow MRI . Liver cirrhosis . Liver haemodynamics . TIPS . Splanchnic system
M. Rössle : M. Schultheiss Department of Gastroenterology, University Medical Center Freiburg, Freiburg, Germany
IVC MELD GRAPPA
M. Markl Department of Biomedical Engineering, McCormick School of Engineering, Northwestern University, Chicago, IL, USA
PC-MRA HVPG
Nitric oxide Endothelin-1 Transjugular intrahepatic portosystemic shunt Doppler Ultrasound Magnetic resonance imaging Time-resolved (CINE) phase contrast (PC) gradient echo sequence with 3-directional and 3-dimensional velocity mapping Inferior vena cava Model for End-Stage Liver Disease Generalized auto-calibrating partially parallel acquisitions 3D phase-contrast MR angiogram Hepatic venous pressure gradient
Eur Radiol
Introduction Symptomatic portal hypertension leads to several clinical symptoms such as variceal bleeding and ascites which often occurs when portosystemic pressure gradients exceed 12 mmHg [1]. It is commonly accepted that increases in portal pressure result from structural and dynamic components. Architectural changes in the liver cause an increase in vascular resistance, the morphological basis of the ‘backward’ component of portal hypertension [2, 3]. It may be augmented by dynamic and reversible constriction of sinusoids by hepatic stellate cells, postulated to be secondary to faulty metabolism of nitric oxide (NO) and imbalanced vasoactive compounds such as endothelin-1 (ET-1), angiotensin II, catecholamines and leukotrienes [4, 5]. A forward flow component has also been postulated. Vasodilating agents such as NO, produced by the congested gut and under the influence of bacterial toxins, escape hepatic clearance and lead to the systemic vasodilatation most evident in the splanchnic blood vessels. This raises the splanchnic blood flow to a degree that augments portal hypertension significantly [6–9]. According to this theory, sinusoidal (structural) hypertension initiates portal hypertension but is itself not the dominant factor. The forward flow theory is based on numerous studies in animal models of various models of cirrhosis, portal vein obstruction and chronic schistosomiasis [5, 10]. Changes in the splanchnic and portal systems are accompanied by a decrease in peripheral resistance and an increase in plasma volume and cardiac output. Most of the clinical haemodynamic alterations in patients with cirrhosis can be explained by simply translating the animal findings to humans. However, reliable measurements in splanchnic arterial flows in humans have not been taken to date because of methodological difficulties in quantifying the flow in splanchnic arterial vessels. To further investigate the haemodynamic alterations in the splanchnic system in patients with chronic portal hypertension, simultaneous measurement of portal and arterial flows may help us prove the forward flow hypothesis and quantify collateral blood flow as the difference between arterial and portal flows in the individual patient. Duplex sonography may not enable us to provide reliable data because measurements of the flow in splanchnic arteries are compromised by the difficulty of measuring vessel diameters. Moreover, duplex sonography strongly depends on the investigator’s skill and is hampered by conditions such as ascites, excessive body weight and bowel gas, all of which restrict its efficacy during quantitative splanchnic haemodynamic assessment. A potential alternative diagnostic imaging approach based on non-contrast four-dimensional (4D) flow MRI enables us to visualize and quantify the blood flow in various regions of the portal and splanchnic arterial systems [11–17], including transjugular intrahepatic portosystemic shunts (TIPS) [18]. The method may also be used to calculate pressure gradients [19, 20]. Thus, 4D flow MRI may allow ‘one-stop-shopping’
to obtain the entire spectrum of information on portal, arterial, collateral and shunt haemodynamics. The purpose of this study was to quantify hepatic arterial and portal haemodynamics in patients with liver cirrhosis using noncontrast 4D flow MRI at 3 T and to investigate changes in hepatic haemodynamics following TIPS insertion.
Materials and methods Study population Written informed consent for participation in this prospective study was obtained from each patient before MRI examination, and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by our local ethics boards. Eleven patients with advanced cirrhosis were included, eight at the University Medical Center, Freiburg, Germany and three at the Northwestern University in Chicago, IL, USA. The study cohort underwent a total of 22 4D flow MRI examinations. All 11 patients were examined before and 4 weeks after TIPS implantation. The aetiologies of cirrhosis were alcohol abuse (n=9), chronic viral hepatitis (n=1) and primary biliary cirrhosis (n=1). Covered stents where implanted preferentially. However, in patients with a high risk of shunt-induced side effects, e.g. older patients, patients with previous episodes of hepatic encephalopathy, patients with bilirubin concentrations of greater than 3 mg/ dl or patients with a Child-Pugh score of >9, bare stents were implanted. As a result four patients received a bare stent (Nitinol Stent System OTW, Eucatech AG, Rheinfelden, Germany; Sinus-SuperFlex-Visual 6 F, Optimed, Ettlingen, Germany), while seven patients received a covered TIPS stent (Viatorr, W. L. Gore & Associates Inc., Flagstaff, AZ, USA). In most of the cases (nine out of 11) the reason for the TIPS placement was refractory ascites, in two cases the reason was variceal bleeding. Ten patients (94 %) received a stent of 10-mm diameter while one obtained a 12-mm stent. Stent length was 6 cm in nine cases (82 %) and 4 cm and 7 cm in one patient each. All measurements were taken after overnight fasting. The patients’ biometrical and biochemical characteristics are shown in Table 1. MR imaging A non-contrast-enhanced three-dimensional (3D) time-resolved (cine) rf-spoiled phase contrast (PC) gradient echo sequence with three-directional velocity encoding (4D flow MRI) was used to assess liver haemodynamics [21, 22]. 4D flow MRI was performed on a 3 T MR system (Magnetom TRIO and Magnetom SKYRA, Siemens, Erlangen, Germany) using a flexible six-element chest coil and six-element spine coil representing the receiver system. All scans were taken during free breathing
Eur Radiol Table 1
Characteristics of the patients before TIPS-implantation Patients (n=11)
Age (years) Sex (female/ male) Body weight (kg) Height (cm) Heart rate (bpm) Blood pressure before MRI (mmhg) Blood pressure after MRI (mmhg) Bilirubin (mg/dl) Creatinine (mg/dl) Albumin (g/dl) INR Refractory ascites (n) Clinical hepatic encephalopathy (n) Child-Pugh class A/B/C (n) Child-Pugh score MELD score
62.5±9.0 3/8 71.6±11.0 171.6±10.8 65.3±6.2 113/68±9/7 113/69±9/6 2.0±1.2 1.0±0.3 3.0±0.4 1.3±0.3 10 0 2/8/1 7.7±1.9 10.9±4.1
with navigator gating at the lung-spleen interface. An enddiastolic navigator signal detected the lung-spleen boundary, which was used for respiratory gating to end-expiration (acceptance window=7 mm) in order to minimize image blurring and ghosting artefacts. 4D flow MRI used prospective ECG-gating combined with k-space segmented data acquisition with Nk =4 phase encoding lines for each cardiac time frame for synchronizing data acquisition during the cardiac cycle. Velocity encoding was obtained by successively executing one reference and three velocity sensitive scans with a temporal resolution of 80 ms and a spatial resolution of 2.4×2.1×2.6 mm3 (TE=2.6 msec, TR= 5 ms). Four-D flow MRI data were acquired in an axial oblique 3D volume angulated along the portal vein with complete volumetric coverage of the hepatic arterial and portal venous systems including the TIPS and outflow in the hepatic veins and inferior vena cava (IVC). Velocity-encoding sensitivity was optimized for the high flow in the hepatic arterial system and TIPS with a value of 100 cm/s in all three flowencoding directions. To shorten the imaging time, we took all scans with k-t GRAPPA acceleration factor R=5. Further parameters were: FOV=260×320 mm2, imaging matrix= 210×160, flip angle α=7°, band width=450 Hz/pixel. The number of cardiac time frames and total imaging time for each examination depended on the efficacy of the respiratory navigator gating and subjects’ heart rates. Data analysis To reduce phase offset errors, Maxwell and eddy current corrections were applied prior to 4D flow data analysis. Maxwell correction was accomplished as described by Bernstein et al. [23] during image reconstruction on the MR system. Data was
subsequently pre-processed using a home-built analysis tool (Matlab, the MathWorks, Natick, MA, USA). This involved velocity anti-aliasing [24], noise filtering and correction for eddy currents as described by Walker et al. [25]. After preprocessing, a 3D phase-contrast (PC) MR angiogram (MRA) was derived from the 4D flow MRI data as described previously [26]. To obtain 3D segmentation of the arterial, portal venous and venous systems, the 3D PC-MRA data was imported to commercial segmentation software (Mimics, Materialise NV, Plymouth, MI, USA) to isolate the velocity data in each vascular region (Fig. 1a). The consequent velocity masks of the 3D PC-MRA were used to depict hepatic arterial and portal venous vascular geometry using iso-surface rendering (EnSight, CEI, Apex, NC, USA). 3D PC-MRA was used to manually place ten analysis planes in the arterial and portal venous systems at the following locations: the splenic and superior mesenteric veins, splenic-mesenteric confluence, proximal portal vein, right and left intrahepatic portal vein branches, coeliac trunk, and splenic, hepatic and superior mesenteric arteries (Fig. 1b). 3D blood flow visualization For 3D flow visualization, all ten analysis planes were used as emitter planes to generate time-resolved 3D particle traces that displayed the temporal dynamics of blood flow over one cardiac cycle [27]. All datasets were evaluated using 3D viewing software (EnLiten, CEI, Apex, NC, USA) enabling visual flow pattern grading by interactive rotation and inspection of the time-resolved particle traces from any view angle. For each of the ten vessels, two independent, experienced readers (Z.S., radiologist, nine years’ abdominal MRI experience; J.C., radiologist, nine years’ cardiovascular MRI experience) semi-quantitatively evaluated image quality of the PC-MRAs and particle traces from the 4D flow MRI datasets. The PCMRA quality grading was based on the vessels’ overall visibility in the portal venous and arterial systems of the liver in all 11 patients before and after TIPS according to a three-point Likert scale (0=not visible, 1=partially visible, 2=completely visible) [28]. Readers were blinded to each other's results, patientidentifying information and the severity of liver disease. 3D blood flow quantification We conducted retrospective quantitative haemodynamic parameter analyses on seven analysis planes (arterial: coeliac trunk, hepatic artery, splenic artery and superior mesenteric artery; portal venous: splenic vein, superior mesenteric vein and portal vein). Net flow over the cardiac cycle and peak velocities were calculated for all 11 patients (baseline and 1 month follow-up examination). The data was acquired during the R-R interval and averaged over multiple cardiac cycles. The 4D flow MRI measurements represent 80–90 % of
Eur Radiol
Fig. 1 4D flow MRI in a 60-year-old male with liver cirrhosis 1 month after TIPS placement. a 3D segmentation of the splanchnic and liver haemodynamics (red: arterial system; blue: venous system; yellow: portal venous system; green: TIPS). abd. aorta=abdominal aorta; spl. artery=splenic artery; spl. vein=splenic vein, hep. artery=hepatic artery;
sma=superior mesenteric artery; smv=superior mesenteric vein; IVC=inferior vena cava. b Colour-coded 3D particle-traces visualization demonstrates increased velocities within the portal vein and TIPS. Ten blue analysis planes were positioned throughout the arterial and portal venous systems to quantify liver haemodynamics
the blood flow within the cardiac cycle including all of the systole and the majority of the diastole. For the constant flow of the portal venous system, corrections are performed based on the acquired time frames extrapolating the constant flow to cover 100 % of the R-R interval. Using 3D segmentation, a home-built analysis tool was applied (Matlab, the MathWorks, Natick, MA, USA) to calculate the peak and mean velocities within the stent. The entire stent was evaluated including the inflow in the stent’s leading end at the portal venous branch through its trailing end within the hepatic vein. In line with the Doppler ultrasound (US) literature, we applied peak shunt velocities below 50 cm/s as cutoff to separate patients with and without shunt stenosis [29, 30]. In addition, we performed invasive hepatic venous pressure gradient (HVPG) measurements before and after TIPS stent graft placement. HVPG represents the gradient between pressure measurements in the intra-abdominal part of inferior vena cava and the portal vein. During the intervention the catheter was inserted into the right jugular vein, passing the right atrium, into the inferior vena cava and into the portal vein.
Results
Statistical analysis Statistical analysis was performed using commercially available software (SPSS 21.0; SPSS, Chicago, IL, USA). Continuous variables are presented as mean±standard deviation. Inter-observer agreement was assessed using Cohen’s kappa statistics with kappa values below 0.20 representing poor agreement, 0.21–0.40 fair agreement, 0.41–0.60 modest agreement, 0.61–0.80 good agreement and 0.81–1.00 excellent agreement [31]. Continuous variables were evaluated by paired, two-tailed t-tests. A p-value
