Quantitative liver perfusion imaging using DCE-MRI has been recently developed to investigate various liver pathologies¹. While recent studies with a dual-input single-compartment model have shown significant differences in perfusion parameters between different types of focal lesions^2,3, it has been suggested that a more appropriate kinetics model for certain pathological settings such as neoplasms utilizes two tissue compartments⁴. This model accounts for the heterogeneity of pathological tissue (which is not composed of just sinusoids), as well as the hepatic artery and portal vein inputs. Thus far, this model has been applied with repeated breath-holds interrupted to allow the patient to breathe. The objective of this study is to test the feasibility of applying this model to 3D free-breathing liver DCE data in the setting of hepatocellular carcinoma (HCC) and metastatic disease, acquired using 3D Through Time Spiral GRAPPA acceleration.

The dual-input two-compartment model has been developed and described in depth by Koh, et al⁴. Key details of the model are given here to familiarize the reader with the model. The contrast agent concentration-time curves in the liver tissue, hepatic artery, and portal vein are represented by C_t(t), C_a(t), and C_p(t), respectively. Blood flow from the hepatic artery and portal vein are represented by F_a and F_p, respectively. F represents the total blood flow to the liver and is equal to the sum of F_a and F_p. $$$ \alpha $$$ is the arterial fraction of the total blood flow to the liver and is equal to F_a/F. All of these quantities are related in the following formula: $$ C_t(t) = \left[ F_aC_a(t) + F_p C_p(t) \right] \ast R(t) = F\left[ \alpha C_a(t) + (1-\alpha)C_p(t) \right] \ast R(t) $$where $$$ * $$$ represents the convolution of the two functions. R(t) is the tissue impulse residue function and is the sum of two functions, R₁(t) and R₂(t), which are represented as$$ R_1(t) = u(t) - u(t-t_1) $$and$$ R_2(t) = u(t) \left\{ 1 - \exp \left(-\frac{PS}{F}\right) \times \left[ 1 + \int_0^t \exp\left(-\frac{PS}{v_2}\tau \right) \sqrt{\frac{PS}{v_2} \frac{PS}{F} \frac{1}{\tau}} I_1 \left(2 \sqrt{\frac{PS}{v_2} \frac{PS}{F}\tau}\right)d \tau\right] \right\}$$u(t) represents the Heaviside unit step function. t₁ represents the mean time for blood to flow through the vasculature. v₂ is the fractional volume of the pathological tissue compartment, PS is the permeability-surface area product of the pathological tissue vasculature, and I₁ is the modified Bessel function. The dual-input two-compartment model was applied to 3D liver perfusion imaging data acquired from a Siemens 3T Skyra scanner. 7 normal volunteers (M:F, 4:3; mean age 20.8 years) and 11 patients (M:F, 9:2; mean age 64.3 years; 6 with metastatic adenocarcinoma with 39 total lesions, 5 with HCC with 13 total lesions) were included in this study. For each subject, high spatiotemporal resolution 3D whole-liver images (1.9×1.9×3.0 mm3; 60 partitions; 1.6 ~ 2 seconds acquisition time) were acquired using the through-time spiral GRAPPA technique with an in-plane reduction factor of 6. A total of 100 to 120 volumes for each subject were acquired continuously in about 4 min, while the subjects were breathing freely.

Fig. 1 shows representative liver perfusion maps obtained from a normal subject and Fig. 2 shows the perfusion maps from a patient with multiple metastatic lesions. A summary of all the perfusion parameters obtained from the 7 normal subjects and 13 patients is presented in Table 1. Compared to results from the normal volunteers, significant differences were observed in F, $$$ \alpha $$$, v₂, PS and t₁ for patients with metastatic adenocarcinoma (P<0.005), and in F, $$$ \alpha $$$, PS and t₁ for patients with HCC (P<0.05). A statistical difference was also observed in F and v₂ between metastases and HCC (P<0.005, Table 1).The dual-input two-compartment model was successfully applied to 3D free-breathing liver perfusion data acquired from both normal subjects and patients with different focal lesions. Significant differences in perfusion parameters were found between normal hepatic parenchyma and focal lesions, and also between different types of lesions. As expected, arterial fraction in both lesion types is higher than in normal parenchyma. Total perfusion, F, is found to be higher in lesions, and is higher in metastatic lesions than in HCC. Fractional volume of the pathological tissue compartment is higher in adenocarcinoma metastases than HCC. We continue to explore the physiological implications of these findings. Further investigation is needed to explore whether a two-compartment model offers advantages in quantitative lesion characterization over a single compartment model. More data will be needed for direct comparison of these techniques. This study demonstrates initial feasibility of utilizing a two-compartment model in 3D free-breathing lesion characterization.