Modelling glioma progression, mass effect and intracranial pressure in patient anatomy

Jana Lipková, Bjoern Menze, Benedikt Wiestler, Petros Koumoutsakos, John LowengrubJournal of The Royal Society Interface 2022 · DOI

TL;DR

This paper introduces a novel computational model that simulates glioma progression, its physical impact on the brain (mass effect), and the resulting intracranial pressure (ICP) changes within patient-specific anatomy. By directly linking tumor dynamics to pressure and deformation, the model offers non-invasive insights into critical conditions and potential neurological impairments, paving the way for improved clinical interventions.

What Problem Does This Paper Address?

Gliomas, a common type of brain tumor, pose a significant threat due to their infiltrative nature and the severe symptoms they cause, often stemming from increased intracranial pressure (ICP). While existing computational models can simulate tumor growth and spread, they largely fail to account for the crucial mechanical interactions between the tumor and the surrounding brain tissue, particularly the resulting pressure distribution. This limitation prevents accurate prediction of ICP-related symptoms like neurological deficits and intracranial hypertension. Furthermore, many existing models rely on a large number of parameters, making them difficult to calibrate with limited patient data. This paper aims to bridge this gap by developing a model that integrates tumor growth, mass effect, and ICP dynamics within patient-specific anatomy, while also simplifying parameterization for better clinical applicability.

Methods

The core of this work is a novel biomechanical model that couples three key processes:

  • Tumor Dynamics: Modeled using a Fisher-Kolmogorov equation that describes tumor proliferation and infiltration into surrounding brain tissue. This accounts for the tumor's growth rate and its ability to spread.
  • Pressure Dynamics: This is a novel aspect. The model treats the growing tumor as a source of pressure on the brain tissue. The brain, modeled as a viscoelastic material, partially relaxes this pressure based on its tissue-specific mechanical properties. The remaining pressure drives deformation. This pressure dynamics is governed by a Helmholtz-like equation, allowing for the computation of the pressure field within the brain.
  • Tissue Dynamics: Brain deformation is modeled using Darcy's law, which relates the deformation rate to the pressure gradient. The model accounts for the displacement and compression of brain tissues due to the tumor's mass effect.

A key innovation is the use of a diffuse domain formulation (DDM). This technique transforms the complex, irregular domain of the brain into a regular domain, simplifying the numerical solution of the governing equations (specifically the Helmholtz equation for pressure). This significantly reduces computational cost and allows for the use of efficient numerical methods, making the model more practical for clinical applications. The implementation is publicly released as open-source software.

Datasets

The study utilizes two types of data:

  • Synthetic Case: A healthy brain atlas (SRI24 Atlas) was used to simulate disease progression. This allows for controlled testing of the model's capabilities without the complexities of real patient data.
  • Clinical Cases: MRI scans from patients diagnosed with High-Grade Glioma (HGG) and Low-Grade Glioma (LGG) were used. These scans were obtained from the Brain Tumor Segmentation (BraTS) challenge dataset, a publicly available benchmark for brain tumor segmentation. The study notes that patient-specific brain anatomy before tumor onset is often unavailable, necessitating the use of affine registration to map an atlas to the patient's scans as an approximation for the initial condition.

Key Insights and Ideas

  • ICP as a Direct Output: The most significant insight is the direct derivation of intracranial pressure from tumor dynamics and patient anatomy. This moves beyond models that only consider pressure as a uniform increment or ignore it altogether, providing a more realistic and localized pressure distribution.
  • Mass Effect as a Physical Process: The model explicitly links tumor growth to physical forces (pressure) that cause brain deformation and displacement. This allows for the prediction of critical conditions like brain midline shift, which is a direct consequence of this physical interaction.
  • Simplified Parameterization: By directly linking pressure to tumor dynamics and anatomy, the model reduces the number of parameters required for calibration compared to previous approaches, making it more amenable to patient-specific modeling.
  • Computational Efficiency through DDM: The adoption of the diffuse domain formulation is a crucial methodological innovation, enabling efficient numerical solutions for complex geometries and reducing the computational burden, which is vital for clinical translation.

Key Results

  • Synthetic Case: The model successfully simulated tumor progression over 600 days, demonstrating realistic brain deformations, including ventricle compression and brain midline displacement. The predicted ICP distribution correlated with the tumor's location and size, and the pressure values were comparable to those associated with pathological conditions. The model also illustrated how pressure elevation in specific brain regions could correspond to potential neurological symptoms.
  • High-Grade Glioma (HGG) Cases: For two HGG patients, the model reproduced the observed mass effect and brain deformations seen in their MRI scans. The simulated pressure distribution closely followed the tumor morphology, and the predicted high ICP values were consistent with intracranial hypertension.
  • Low-Grade Glioma (LGG) Case: The model was tested on a patient with scans at two time points, showing progression from LGG to HGG. It accurately captured the initial mild mass effect of the LGG and the more significant compression and midline shift associated with the HGG stage.

Limitations and Future Work

  • CSF Circulation: The model does not account for cerebrospinal fluid (CSF) circulation. This limitation means it cannot capture phenomena like ventricular enlargement due to CSF changes, which can occur in some patients. Future work should incorporate CSF dynamics.
  • Brainstem Deformation: The model does not explicitly account for distortions of the brainstem, which can be significant in advanced glioblastoma.
  • Anisotropic Infiltration: While the model considers tissue-dependent infiltration, it doesn't explicitly model anisotropic tumor cell migration along white matter fibers, which could be informed by diffusion tensor imaging (DTI). However, challenges with DTI signal corruption in oedematous regions and fiber orientation changes during deformation are noted.
  • Tumor Compression: The current model assumes tumors are not compressed by surrounding tissue, which is generally true for infiltrative gliomas. However, for other tumor types or specific scenarios, this might need to be addressed.
  • Model Calibration: While simplified, further validation with a larger patient cohort is needed to fully assess the model's calibration and accuracy for patient-specific predictions.
  • Post-operative Changes: The model currently focuses on tumor-induced effects. Extending it to simulate post-operative scenarios, including tissue decompression and the role of residual stresses, is a promising future direction.

The authors emphasize that the publicly released implementation aims to facilitate clinical deployment and further research in this critical area.

Summary updated

Abstract

Increased intracranial pressure is the source of most critical symptoms in patients with glioma, and often the main cause of death. Clinical interventions could benefit from non-invasive estimates of the pressure distribution in the patient's parenchyma provided by computational models. However, existing glioma models do not simulate the pressure distribution and they rely on a large number of model parameters, which complicates their calibration from available patient data. Here we present a novel model for glioma growth, pressure distribution and corresponding brain deformation. The distinct feature of our approach is that the pressure is directly derived from tumour dynamics and patient-specific anatomy, providing non-invasive insights into the patient's state. The model predictions allow estimation of critical conditions such as intracranial hypertension, brain midline shift or neurological and cognitive impairments. A diffuse-domain formalism is employed to allow for efficient numerical implementation of the model in the patient-specific brain anatomy. The model is tested on synthetic and clinical cases. To facilitate clinical deployment, a high-performance computing implementation of the model has been publicly released.
BibTeX
@article{lipkova2022modelling,
  author = {Jana Lipková and Bjoern Menze and Benedikt Wiestler and Petros Koumoutsakos and John Lowengrub},
  title = {{Modelling glioma progression, mass effect and intracranial pressure in patient anatomy}},
  journal = {Journal of The Royal Society Interface},
  year = {2022},
  doi = {10.1098/rsif.2021.0922},
}
Figure