The separation of portal-venous and hepatic arterial blood supply is important for the evaluation of various liver diseases, especially for the diagnosis of hepatocellular carcinoma.¹ Arterial spin labeling (ASL) has been shown to be a promising non-invasive approach for perfusion measurement in various organs such as the brain or the kidney.^2,3 A small number of studies have reported the application of ASL for measurement of liver perfusion in humans.^4,5,6,7 However, liver perfusion imaging with ASL has been limited due to respiratory and cardiac motion artifacts. It was demonstrated that background suppression (BS) reduced the sensitivity of ASL measurements to physiological motion.⁸ Furthermore, the separation of arterial and portal-venous perfusion fractions remains difficult, which is caused by the complex geometry and physiology of blood supply to the liver. The aim of this study was to investigate the capability of pseudo-continuous ASL (pCASL) with BS to separate arterial and portal-venous perfusion of the liver by varying the orientations of the tagging plane.

Three healthy volunteers were examined on a 3T MR scanner (MAGNETOM Prisma, Siemens Healthcare) with body and spine array coils. The perfusion of the liver was measured using a pCASL echo planar imaging sequence.⁹ Four to six sagittal slices were acquired with parameters: TR/TE, 8000/15 ms; slice thickness/gap, 8/4 mm; in-plane resolution, 3×3 mm²; acquisition matrix, 86×106; readout bandwidth, 2360 Hz/Pixel. Tagging flip angles (FA) of 30°-35° and a gradient strength of 7 mT/m were used. Post-labeling delay (PLD) and tag duration (TD) were set to 2000 ms. 36 label-control image pairs were acquired within approx. 5 min. BS for liver tissue (T1=800 ms) was utilized using a double inversion approach. Images were acquired by employing a timed breathing protocol.

One approach to separate portal-venous and arterial perfusion is by directly tagging the hepatic artery (HA) and portal vein (PV) blood supply in two separate measurements. For the labeling of HA the tagging plane was placed in an axial facial, cranial to the liver (Figure 2, plane 4), and PV was tagged using a plane below HA (plane 3). Another approach is based on labeling of the total liver and PV blood flow, extracting the arterial perfusion by subtraction. For labeling of total liver perfusion, the tagging plane was placed perpendicular to the PV or parallel to the aorta (planes 1, 2). An additional measurement with a tagging plane outside the body (plane 5) was performed as control.

Any images which showed strong motion artefacts were discarded from subsequent analysis using Matlab. Data from averaged difference images were used to create a mask of liver parenchyma, by segmenting those voxels whose signal was 2 times larger than the median value (corresponding to PV, its branches and hepatic veins). For an estimation of quantitative perfusion values, the mean perfusion over the liver parenchyma mask was calculated using a kinetic model.¹⁰

Figure 1 shows an example of perfusion-weighted images measured with and without BS. Images with BS show clearly higher image quality with reduced motion artifacts. In Figure 2, mean difference images of total liver perfusion as well as portal-venous and arterial perfusion can be seen. Perfusion maps of total liver and PV blood flow are shown in Figure 3. In Figure 4, perfusion maps for total liver and HA blood flow are depicted. Mean perfusion values for total liver blood flow were approx. 83 and 91 ml/min/100g for tagging planes 1 and 2, respectively. Mean perfusion values for PV and HA blood flow were approx. 54 and 13 ml/100g/min, respectively. Perfusion images of total liver blood flow revealed a more homogenous distribution than PV blood flow (Figures 2, 3). The measurement with the tagging plane outside the body resulted in approx. 8 ml/min/100g.In the present work, significant benefits of the background suppression technique for ASL imaging of the liver could be demonstrated. The combination of the pCASL sequence with BS provides high quality perfusion images of the liver and allows for separation of portal-venous and arterial blood supply. Our results show that direct measurement of relatively low arterial perfusion signal could be confined by background noise. Nevertheless, this approach can be advantageous in patients with highly vascularized liver lesions. The extraction of the arterial perfusion fraction from measurements of total liver and PV perfusion appears to be a more robust procedure, however, this approach requires two separate measurements. Both approaches are worth being further refined to find optimal labeling strategies for measurement of liver perfusion.