Hepatic arterial spin labelling MRI: an initial evaluation in mice

Abstract

The development of strategies to combat hepatic disease and augment tissue regeneration has created a need for methods to assess regional liver function. Liver perfusion imaging has the potential to fulfil this need, across a range of hepatic diseases, alongside the assessment of therapeutic response. In this study, the feasibility of hepatic arterial spin labelling (HASL) was assessed for the first time in mice at 9.4 T, its variability and repeatability were evaluated, and it was applied to a model of colorectal liver metastasis. Data were acquired using flow-sensitive alternating inversion recovery-arterial spin labelling (FAIR-ASL) with a Look-Locker readout, and analysed using retrospective respiratory gating and a T1 -based quantification. This study shows that preclinical HASL is feasible and exhibits good repeatability and reproducibility. Mean estimated liver perfusion was 2.2 ± 0.8 mL/g/min (mean ± standard error, n = 10), which agrees well with previous measurements using invasive approaches. Estimates of the variation gave a within-session coefficient of variation (CVWS) of 7%, a between-session coefficient of variation (CVBS) of 9% and a between-animal coefficient of variation (CVA) of 15%. The within-session Bland-Altman repeatability coefficient (RCWS) was 18% and the between-session repeatability coefficient (RCBS) was 29%. Finally, the HASL method was applied to a mouse model of liver metastasis, in which significantly lower mean perfusion (1.1 ± 0.5 mL/g/min, n = 6) was measured within the tumours, as seen by fluorescence histology. These data indicate that precise and accurate liver perfusion estimates can be achieved using ASL techniques, and provide a platform for future studies investigating hepatic perfusion in mouse models of disease.

Keywords: ASL; liver; metastasis; mouse; perfusion; preclinical; repeatability; variability.

Figure 1

Schematic diagram of the respiratory-triggered, Look–Locker T₁ mapping sequence with a spoiled, single-slice, gradient-echo readout. The inversion pulse is end-expiration triggered, followed by a free-breathing, segmented, Look–Locker sampling train. Each segmented block is separated by TI_{Look–Locker}; each block contains four sampling pulses with TR_RF. The sequence is performed twice with differing inversion slice thicknesses as part of the flow-sensitive alternating inversion recovery-arterial spin labelling (FAIR-ASL) design; a hepatic ASL (HASL) dataset is completed in 15 min.

Figure 2

The phase-encoded noise-based image rejection (PENIR) scheme for retrospective removal of motion-corrupted data. (A) Axial image of a mouse liver with an example user-determined region of interest (ROI) (broken line) of the extracorporeal area in the phase-encoding direction. (B) Significant increases in the mean noise signal (blue line) above a signal-generated threshold (red line) will result in image omission (shaded areas). (C) Inversion recovery fitting in a liver region will typically overestimate T₁ without correction (broken line) compared with rejection (full line) because of reduced tissue signal from corrupted images (filled diamonds). (D) A comparison of T₁ maps shows that, with no correction, the respiration artefacts may lead to unsuccessful fitting and generate areas of signal dropout (arrow). With both data-logger (DL) and PENIR retrospective gating, these areas of dropout are recovered – no significant differences in mean T₁ were measured between these two retrospective conditioning modes (t-test, p > 0.05).

Figure 3

Quantification of perfusion from a hepatic arterial spin labelling (HASL) dataset. (A) The perfusion signal comes from the difference in longitudinal recovery following slice-selective (red) and global (blue) inversion. Example T₁ maps acquired after a slice-selective inversion (B) and global inversion (C); T₁ values have been clipped at 2 s (mean liver global T₁ = 1.36 ± 0.06 s). (D) In the resulting perfusion map, the liver parenchyma has been overlaid on a high-resolution T₂-weighted fast spin echo image. Non-physiologically high perfusion values can be seen within major blood vessels; perfusion values in this image have been limited at 10 mL/g/min. Mean estimated liver perfusion was 2.2 ± 0.8 mL/g/min (mean ± standard error, n = 10).

Figure 4

(A) The variability in mean liver perfusion across all 10 animals (shown in different colours). The second imaging session was 1 week later. Across the group, no significant trends were observed within and between the imaging sessions. Bland–Altman plots showing the within-session repeatability from the first and second mean liver perfusion estimate (B) and the between-session mean perfusion repeatability from index-matched scans (C). The central, thick full line represents the mean difference, and the two outer lines show the ±1.96 × standard deviation. For both plots, the mean difference measured was within the error of the technique, and no trends could be distinguished.

Figure 5

Application of hepatic arterial spin labelling (HASL) to a model of liver metastasis. (A) High-resolution T₂-weighted fast spin echo images show the tumours (arrow) as hyperintense relative to the liver tissue (outlined). (B) In the corresponding slice-selective T₁ map, the metastases can be delineated by a raised T₁ (2.24 ± 0.54 s) relative to liver tissue. (C) Across n = 6 mice, a significant reduction in perfusion (outlined) was measured within the metastases (1.1 ± 0.5 mL/g/min, mean ± standard deviation) relative to the liver tissue. This difference in perfusion measured by HASL was confirmed by histology. (D) The fluorescence image shows a section of normal liver containing sinusoid vessels (arrow), demarcated from an adjacent SW1222 colorectal liver metastasis (star) by a large reduction in the presence of blood vessels. Vascular structures are shown in red (anti-CD31 antibody) and perfusion in blue (by the injected marker Hoechst 33342).