Non-invasive Assessment of Hepatic Venous Pressure Gradient (HVPG) Based on MR Flow Imaging and Computational Fluid Dynamics

Kexin Wang, Shuo Wang, Minghua Xiong, Chengyan Wang, He WangMICCAI 2021 · DOI

TL;DR

This paper presents a novel, non-invasive method for measuring the Hepatic Venous Pressure Gradient (HVPG) using a combination of advanced MRI techniques and computational fluid dynamics (CFD). By overcoming previous limitations in hepatic vessel imaging and CFD modeling, this approach offers a promising alternative to invasive clinical measurements for diagnosing clinically significant portal hypertension.

What Problem Does This Paper Address?

Clinically significant portal hypertension (CSPH) is a serious complication of chronic liver disease, often associated with cirrhosis. The current gold standard for diagnosing CSPH is the measurement of the Hepatic Venous Pressure Gradient (HVPG). However, this procedure is invasive, involving the insertion of a catheter, which carries risks of infection and other complications, making it inaccessible or undesirable for some patients. Existing non-invasive methods like transient elastography (TE) have limitations in accuracy, especially in certain patient populations. Therefore, there is a critical need for a reliable, non-invasive method to assess HVPG.

Methods

The authors propose a "one-stop solution" that integrates two key technologies:

  • Multi-contrast MRI: To obtain detailed anatomical information of the hepatic vessels without the need for contrast agents, they employ multi-echo DIXON (mDIXON) imaging. This technique captures both water and fat signals, allowing for the reconstruction of vessel anatomy. For velocity measurements, they use phase-contrast MRI (PC-MRI) with cardiac and respiratory gating to acquire blood flow velocity data.
  • Computational Fluid Dynamics (CFD): The anatomical and velocity data from MRI are then used to build a patient-specific CFD model. The Navier-Stokes equations are solved using a finite element method (FEM) to simulate blood flow and pressure distribution within the hepatic vasculature. A key innovation here is an improved CFD pipeline that addresses challenges in reconstructing vessel connections and reducing computation time. Specifically, they developed a programming method to automatically connect the portal and hepatic veins, creating a more realistic vascular network for the simulation. To mimic the clinical measurement of HVPG, a "break point" is introduced in the model to simulate the blockage of blood flow, allowing for the calculation of virtual wedged hepatic venous pressure (vWHVP) and virtual free hepatic venous pressure (vFHVP). The virtual HVPG (vHVPG) is then calculated as the difference between these two simulated pressures.

Datasets

The study utilized data from:

  • Ten healthy volunteers
  • Five patients with liver cirrhosis (diagnosed via liver biopsy)

The MRI data were processed using pre-trained deep neural networks (DNNs) that were trained on an additional dataset of 30 manually labeled cases from previous work.

Key Insights and Ideas

  • Bridging the MRI-CFD Gap: The paper successfully demonstrates how to integrate detailed MRI-derived vascular anatomy and flow dynamics into a CFD model for a specific organ system (the liver), which has been a significant challenge due to the complex vascular structure and difficulty in obtaining accurate boundary conditions.
  • Automated Vessel Connection: The novel programming approach for connecting portal and hepatic veins is crucial. It simplifies the complex process of creating a continuous vascular network for CFD, making the method more robust and less reliant on manual intervention.
  • Mimicking Invasive Measurement: The introduction of a simulated "break point" in the CFD model to represent the wedged catheter is a clever way to directly replicate the clinical HVPG measurement, allowing for a direct comparison and validation of the non-invasive vHVPG.
  • Rigid Wall Assumption: The authors justify using a rigid wall model for the vessels, acknowledging that while Fluid-Structure Interaction (FSI) would be more accurate, the reduced compliance in patients with CSPH makes the rigid wall assumption computationally efficient and reasonably accurate.

Key Results

  • The proposed MRI-based CFD method successfully simulated pressure gradients in both healthy volunteers and patients with liver cirrhosis.
  • The simulated pressure distribution showed a gradual decrease from the portal vein inlet to the hepatic vein outlet, consistent with physiological expectations.
  • In the tested cases, the vHVPG values correlated well with clinical diagnoses. For example, a healthy volunteer had a vHVPG of 4.7 mmHg, while a patient with liver disease had a vHVPG of 10.2 mmHg, which aligns with the clinical definition of CSPH (HVPG > 10 mmHg).
  • A table summarizing the vHVPG for different groups showed an average vHVPG of 5.9 ± 2.3 mmHg for healthy volunteers and 21.5 ± 8.8 mmHg for patients with liver cirrhosis.
  • The study reported substantial consistency (kappa test, k=0.7) between the diagnosis based on histopathology and the virtual HVPG measurements.
  • The method is computationally efficient, with the current iteration taking less than twenty minutes to compute HVPG from raw images.

Limitations and Future Work

  • Resolution Limitations: The spatial resolution of the MRI (0.87 x 0.87 x 2.0 mm³) limits the visualization of very fine capillaries. While the authors believe this is comparable to conventional CT and adequate for CFD, they acknowledge that the absence of fine-scale capillaries might influence specific wall variables like shear stress.
  • Simplified Physics: The current model uses laminar physics and a rigid wall assumption. The authors suggest that future work should consider more complex fluid dynamics (e.g., turbulence) and Fluid-Structure Interaction (FSI) to create a more precise model.
  • Limited Sample Size: The study included a relatively small number of participants (10 healthy, 5 patients), and the authors note that the number of cases is limited.
  • Future Directions: The authors plan to incorporate the interaction between vessel structures and fluid flow (FSI) in future models. They also aim to further improve the automated construction of vascular networks to minimize manual interference.

Summary updated

BibTeX
@inproceedings{wang2021non,
  author = {Kexin Wang and Shuo Wang and Minghua Xiong and Chengyan Wang and He Wang},
  title = {{Non-invasive Assessment of Hepatic Venous Pressure Gradient (HVPG) Based on MR Flow Imaging and Computational Fluid Dynamics}},
  booktitle = {International Conference on Medical Image Computing and Computer-Assisted Intervention},
  year = {2021},
  doi = {10.1007/978-3-030-87234-2\_4},
}
Figure