Use of Caval Subtraction 2D Phase-Contrast MR Imaging to Measure Total Liver and Hepatic Arterial Blood Flow: Preclinical Validation and Initial Clinical Translation

Caval subtraction phase-contrast MR imaging is technically feasible and may offer a reproducible and clinically viable method for measuring total liver blood flow and hepatic arterial flow.


Purpose:
To validate caval subtraction two-dimensional (2D) phasecontrast magnetic resonance (MR) imaging measurements of total liver blood flow (TLBF) and hepatic arterial fraction in an animal model and evaluate consistency and reproducibility in humans.

Materials and
Methods: Approval from the institutional ethical committee for animal care and research ethics was obtained. Fifteen Sprague-Dawley rats underwent 2D phase-contrast MR imaging of the portal vein (PV) and infrahepatic and suprahepatic inferior vena cava (IVC). TLBF and hepatic arterial flow were estimated by subtracting infrahepatic from suprahepatic IVC flow and PV flow from estimated TLBF, respectively. Direct PV transittime ultrasonography (US) and fluorescent microsphere measurements of hepatic arterial fraction were the standards of reference. Thereafter, consistency of caval subtraction phasecontrast MR imaging-derived TLBF and hepatic arterial flow was assessed in 13 volunteers (mean age, 28.3 years 6 1.4) against directly measured phase-contrast MR imaging PV and proper hepatic arterial inflow; reproducibility was measured after 7 days. Bland-Altman analysis of agreement and coefficient of variation comparisons were undertaken.
TECHNICAL DEVELOPMENT: Measurement of Total Liver and Hepatic Arterial Blood Flow Chouhan et al infrahepatic, suprarenal portion of the IVC (Q infrahepatic ) (Fig E1a [online]), as follows: Hepatic arterial flow (Q HA ) can then be estimated by subtracting directly measured PV flow (Q PV ) from the outflow TLBF ( Fig E1b [online], as follows: The study cohort consisted of 15 healthy animals treated with sham laparotomy (n = 11) or bile duct ligation (BDL) (n = 4). The latter group was included to test the feasibility of performing phase-contrast MR imaging in an imaging is an established, widely available, and validated method for the noninvasive measurement of largevessel bulk flow (2)(3)(4)(5). Several investigators have used phase-contrast MR imaging to study PV flow, although few have reported data from the proper hepatic artery (6-9) because evaluation is challenging in routine clinical practice, primarily because of small vessel size, tortuosity, and anatomic variation. Furthermore, low signalto-noise ratio, vessel orthogonality, partial volume averaging errors, intravoxel phase dispersion, and spatial misregistration can confound estimation of pulsatile flow in small arteries and have impeded both research and clinical use of phase-contrast MR imaging for the measurement of total liver blood flow (TLBF).

Preclinical Validation
Recognizing these challenges, we propose an alternative application of 2D phase-contrast MR imaging to measure TLBF and proper hepatic arterial flow by using caval subtraction. TLBF and hepatic arterial flow fraction could be valuable hemodynamic biomarkers of liver disease. The purpose of this study was to validate caval subtraction 2D phase-contrast MR imaging measurements of TLBF and hepatic arterial fraction in a preclinical animal model and evaluate consistency and reproducibility in humans.

Background Theory
From the application of the principle of conservation of mass to flow (Q), for a fixed tissue volume: where Q in is blood flow into the organ and Q out is blood flow out of the organ. On the basis of the anatomic configuration of the liver, the intrahepatic inferior vena cava (IVC) receives blood entirely from the hepatic venous system. Outflow TLBF (Q out , equivalent to Q TLBF ) can therefore be estimated by measuring bulk flow in the suprahepatic subcardiac portion of the IVC (Q suprahepatic ) and then subtracting flow from the  n Caval subtraction 2D phase-contrast MR imaging is a noninvasive, simple, and rapid technique for measuring total liver and hepatic arterial blood flow.

Advances in Knowledge
T he dual portal venous (PV) and hepatic arterial blood supply to the liver makes it difficult to assess hemodynamic complications of liver disease. To date, invasive methods still remain the standard of reference. Imaging-based hepatic hemodynamic assessment has the potential to yield useful and meaningful biomarkers of portal hypertension and chronic liver disease (1).

TECHNICAL DEVELOPMENT: Measurement of Total Liver and Hepatic Arterial Blood Flow
Chouhan et al basis of established algorithms (12). All PV flow, estimated TLBF, and hepatic arterial flow measurements were normalized to explanted liver weight.

Human Translation
Two aspects of the caval subtraction technique were then tested in healthy volunteers: (a) consistency with directly measured phase-contrast MR imaging PV and hepatic arterial inflow and (b) measurement reproducibility after 7 days. Subjects and preparation.-University College London Ethics Committee approval was obtained, and participants provided informed written consent. Volunteers were recruited by means of advertisement within the university campus and were eligible if they (a) had no contraindications to MR imaging, (b) were not taking any long-term medication (excluding the oral contraceptive pill), and (c) had no documented history of previous liver or gastrointestinal disease. Fourteen volunteers were screened, one of whom was excluded because of claustrophobia. The final cohort consisted of seven men (mean age 6 standard deviation, 26.5 years 6 1.36) and six women (mean age, 31.2 years 6 2.62) who underwent imaging between June 14 and July 25, 2013. Participants fasted for 6 hours before MR imaging and avoided caffeinated fluids.
were used to identify the IVC and PV, and phase-contrast image planes were planned to ensure orthogonality with the subcardiac portion of the suprahepatic IVC, suprarenal portion of the infrahepatic IVC, and PV.
Cardiac-and respiratory-gated 2D cine phase-contrast MR imaging was performed with 2-mm-thick sections, a 10° flip angle, and a 192 3 192 (frequency encoding 3 phase encoding) acquisition matrix. On the basis of initial pilot work in four animals (not presented in this study), velocity encoding (V enc ) settings of 33 cm/sec for PV and infrahepatic IVC flows and 66 cm/ sec for suprahepatic IVC flows were applied (Table). Phase maps were acquired at each V enc setting with opposite flow-encoding directions. Correction for background phase errors was achieved by subtracting phase maps with opposing flow-encoding directions, with the assumption that the phase of stationary spins was identical in each image. Acquisition time for each phasecontrast MR imaging measurement was usually less than 10 minutes. Regions of interest were positioned manually by the study coordinator on each vessel for each frame of the cardiac cycle, and flow quantification was performed by using in-house-developed Matlab code (MathWorks, Natick, Mass) on the animal model of portal hypertension. BDL and sham surgery were conducted as described previously (10). After recovery, animals were maintained for 4-5 weeks before undergoing the experimental protocol. All procedures were performed by the study coordinator (M.D.C., a radiology research fellow qualified in animal handling with 4 years of experience).
Standards of reference.-For rats that underwent sham operation (n = 11), laparotomy was performed and a 2-mm transit-time US probe (Transonic Systems, Ithaca, NY) was placed around the PV. PV flow readings were obtained with transit-time US after 10-15 minutes, once the animal was stable. Extensive adhesions around the porta hepatis and associated high risk of traumatic vessel injury precluded transit-time US validation in the BDL group.
After transit-time US measurement, 15-mm polystyrene fluorescent microspheres (FluoSpheres; Life Technologies, Warrington, England) suspended in heparinized saline were administered transcutaneously into the left ventricle over approximately 10 seconds under US guidance (Terason; Teratech, Burlington, Mass) ( Fig E2  [online]). Animals were then transferred to the imager for phase-contrast MR imaging. The animal was then sacrificed and organs were explanted for microsphere processing using an adapted protocol (11), as summarized in Appendix E1 (online). To ensure adequate central mixing of microspheres, data exceeding 20% difference in microsphere content between right and left kidneys were excluded.
Two-dimensional cine phase-contrast MR imaging.-Temperature was monitored with a rectal probe (SA Instruments, New York, NY), with core body temperature maintained between 36°C and 38°C. Cardiac monitoring was undertaken by using a triple-electrode single-lead system (SA Instruments). Imaging was performed by using a 9.4-T unit (Agilent Technologies, Oxford, England), with sequence parameters listed in the  (Fig E3  [online]).
(Eqq [2,3]) and compared with direct phase-contrast MR imaging of PV and hepatic arterial inflow.
Liver volume was estimated by using steady-state free precession coronal images with 5-mm-thick sections. Segmentation was performed manually by the study coordinator using software (Amira Resolve RT; Visage Imaging, Berlin, Germany). A tissue density of 1.0 g/mL was assumed (13).
Seven days after the original study, subjects underwent repeat imaging with identical preparation and MR imaging protocol at a comparable time of the day. All analyses and quantification were performed by the study coordinator (M.D.C., with 5 years of experience in abdominal imaging).

Statistical Analysis
Data normality was confirmed with Kolmogorov-Smirnov testing. All bulk flow measurements obtained at phasecontrast MR imaging were normalized to liver weight or volume. Comparison between measurements derived from caval subtraction phase-contrast MR imaging and standards of reference (transit-time US, microspheres, direct inflow phase-contrast MR imaging) and 7-day reproducibility studies were assessed by using Bland-Altman analysis of agreement, with calculation of 95% limits of agreement (LOA). Coefficients of variation were also calculated and compared by using methods described by Forkman (14). Because of the small number of animals in the sham and BDL groups that underwent validation with microsphere analysis, validation analysis was pooled across both cohorts and the Mann-Whitney U test used for comparison of PV and relative hepatic arterial flow. Data are expressed as means 6 standard errors, and P , .05 was indicative of a statistically significant difference.

Preclinical Cohort
Across 15 animals, the mean body weight was 451.7 g 6 9.0 and the wet liver mass was 24.7 g 6 2.0. All animals Two-dimensional cine phase-contrast MR imaging.-Phase-contrast MR imaging was performed with a 3.0-T unit (Achieva; Philips Healthcare, Best, the Netherlands) and a 16-channel body coil (SENSE XL-Torso, Philips Healthcare). Imaging parameters are given in the Table. Coronal (upper abdomen), sagittal (abdominal great vessels), and oblique (PV) breath-hold balanced steadystate free precession images were acquired. Two-dimensional phasecontrast MR imaging with expiratory breath hold and retrospective cardiac gating was planned in two planes by the study coordinator to ensure orthogonality to the target vessel. Studies were performed through the PV (V enc = 40 cm/sec), proper hepatic artery (V enc = 60 cm/sec), infrahepatic IVC (above the renal veins, below the hepatic IVC; V enc = 60 cm/sec), and suprahepatic IVC (above the hepatic venous inflow, below the right atrial junction; V enc = 80 cm/sec). Where hepatic arterial anatomy varied (n = 2), measurements were made as close as possible to the origin of the hepatic artery. Images were reviewed for aliasing and V enc settings increased by 20 cm/sec when appropriate. Data were acquired by using the unit's clinical flow quantification implementation. Phase maps were acquired at each V enc setting with opposite flowencoding directions. Correction for background phase errors was achieved by subtracting phase maps with opposing flow-encoding directions, with the assumption that the phase of stationary spins was identical in each image. A local phase-correction filter was also applied to correct for phase errors induced by eddy currents. The acquisition time for each measurement was less than 20 seconds. Each phase-contrast MR imaging study was repeated three times. Flow quantification was performed by using freely available software (Segment; Medviso, Lund, Sweden) and the mean of triplicate measurements used for analysis. Caval subtraction TLBF, PV flow, hepatic arterial flow, and hepatic arterial fraction were calculated TECHNICAL DEVELOPMENT: Measurement of Total Liver and Hepatic Arterial Blood Flow Chouhan et al min/100 g, respectively) (Fig 2a, 2c). The coefficient of variation for caval subtraction phase-contrast MR imaging TLBF (23.0%) was similar to that for direct inflow TLBF (22.2%) (F 23,23 = 0.94, P = .56). Hepatic arterial flow with caval subtraction phase-contrast MR imaging (123.2%) was higher than that with direct phase-contrast MR imaging (75.6%), although this did not reach statistical significance (F 23,23 = 1.68, P = .11). Seven-day reproducibility.-The mean differences between TLBF measured with caval subtraction phasecontrast MR imaging and hepatic arterial flow measurements obtained 7 days apart were 28.5 mL/min/100 g 6 4.9 TLBF measured with caval subtraction phase-contrast MR imaging (mean, 71.1 mL/min/100 g 6 3.3) and that measured with direct phase-contrast MR imaging (sum of PV and common hepatic arterial flow; mean, 72.5 mL/min/100 g 6 3.3) and between calculated hepatic arterial flow (Eq [3]; mean, 14.0 mL/min/100 g 6 3.5) and direct phase-contrast MR imaging measured hepatic arterial flow (mean, 15.3 mL/min/100 g 6 2.4) was 21.3 mL/min/100 g 6 2.4. The 95% LOA for caval subtraction versus direct inflow phase-contrast MR imaging were 623.1 mL/min/100 g for both TLBF and hepatic arterial flow (range, 40.5-100.9 mL/min/100 g and 220.7 to 54.9 mL/ Clinical Cohort The mean liver volume was 1211.0 mL 6 52.9. Two subjects declined subsequent examination, leaving 11 in the reproducibility cohort. Technical feasibility of caval subtraction phase-contrast MR imaging.-Electrocardiographically and respiratory-gated cine phase-contrast MR imaging flow studies through the cardiac cycle demonstrated physiologic flow profiles through the PV, infrahepatic IVC, and suprahepatic IVC ( Fig  E4 [online]).
Caval subtraction phase-contrast MR imaging versus direct phase-contrast MR imaging.-The mean difference between  . Furthermore, frequent anatomic variation (18)(19)(20) complicates measurement and requires costly radiologic expertise for planning, which substantially impedes clinical implementation. We have demonstrated that caval subtraction phase-contrast MR imaging provides a relatively simple strategy for overcoming these limitations in both rodents and humans. We found that phase-contrast MR imaging is technically feasible in animals (even in a cirrhotic BDL model), and phase-contrast MR imaging PV flow estimates showed

Discussion
Accurate assessment of proper hepatic arterial flow (and, consequently, TLBF) with use of phase-contrast MR imaging is challenging in both preclinical and clinical contexts. In rodents, the vessel itself is extremely small-less than 1 mm in diameter (15)(16)(17)-which introduces measurement errors, even at high field strengths. Technical challenges have confounded attempts to measure proper hepatic arterial flow in humans with clinical systems (6)(7)(8)(9). Adequate signal-to-noise ratio is less of an issue because the vessel is larger, but partial and 7.3 mL/min/100 g 6 4.4, respectively. The 95% LOA were 631.6 mL/ min/100 g for TLBF and 628.8 mL/ min/100 g for hepatic arterial flow. The mean difference between directly measured phase-contrast MR imaging TLBF and hepatic arterial flow measurements obtained 7 days apart were 22.3 mL/ min/100 g 6 4.5 and 1.1 mL/min/100 g 6 3.0, respectively. The caval subtraction 95% LOA were 631.6 mL/min/100 g for TLBF and 628.8 mL/min/100 g for hepatic arterial flow. The 95% LOA were 629.6 mL/min/100 g for TLBF and 619.5 mL/min/100 g (range, 4.9-44.6 mL/min/100 g) for hepatic arterial flow.     (21). Thereafter, with the caval subtraction technique, reasonable agreement was obtained for hepatic arterial fraction against a microsphere standard of reference. Unfortunately, quantification of absolute hepatic arterial flow was not undertaken because simultaneous peripheral arterial sampling proved unreliable in pilot experiments.
For human translation, we chose not to use transcutaneous Doppler US measurements of PV and hepatic arterial flow because reported reproducibility is poor (8,(22)(23)(24). Instead, direct phasecontrast MR imaging measurements of PV and hepatic arterial flow were used to test consistency of the caval subtraction phase-contrast MR imaging technique. Good agreement for TLBF was demonstrated between the two methods, and although hepatic arterial flow agreement was less impressive, the level of disagreement was not contingent on measurement value (ie, there was no systematic bias). Hepatic arterial flow measurements suffer from error propagation as inaccuracies in IVC and PV flow are summated during subtraction. This can result in nonphysiologic results such as negative estimates of hepatic arterial flow (particularly when true hepatic arterial flow is low, as would be expected in healthy volunteers).
Measurements obtained with caval subtraction phase-contrast MR imaging were, however, reassuringly similar to the direct measurements obtained with inflow phase-contrast MR imaging, although variable for both techniques, likely due to natural variation in vessel flow rates contingent on differing subject hydration, for example.
To place the levels of agreement for caval subtraction technique with standards of reference into clinical context, it is known that TLBF can vary by as much as 58% between health and disease (16,25,26), which is much greater than the expected error range found in the current study.
Our study has limitations. Although hepatic venous pressure gradient and portal venous pressure are clinically useful hepatic hemodynamic