GlimS investigated mathematical models that capture two dominant characteristics of macroscopic glioma growth: healthy-tissue infiltration and tissue-displacing mass-effect. These models allow simulating the spatio-temporal evolution of tumor growth with patient-specific parameters in a computational representation of the human brain. GlimS evaluated models with and without mass-effect, and combined them with approaches to solve the inverse problem: to identify the model parameters that best reproduce the imaging data of a patient's tumor. Mathematical models, image processing algorithms, and parameter estimation approaches were implemented using open-source libraries. The resulting software is available on the project website and enabled the study of specific questions linked to the role of biomechanics in the mathematical modeling of this tumor:
First, GlimS investigated the significance of directional (anisotropic) brain tissue characteristics for the formation of tumor shapes in mathematical growth models. Brain tissue is spatially heterogeneous. Besides, its microscopic organization gives rise to spatial structures that influence the direction of tumor cell migration. In an in-silico study, GlimS examined how this direction-dependence of tissue properties affects tumor shape. The results indicate that tissue anisotropy is not a major determinant of macroscopic tumor shape in most growth locations.
Second, GlimS evaluated image-derived surrogate measures of tumor mass-effect. Tumor mass-effect is poorly quantified in clinical practice. The most common quantitative measure, Brain midline shift (MLs), is determined from 2D image slices. Lateral ventricle displacement (LVd) is an alternative metric that accounts for the entire 3D anatomy of the brain. Comparing MLs and LVd in an in-silico cohort study revealed LVd as the more robust and predictive measure. It is insensitive to tumor location, highly correlates with tumor volume, and is a good predictor of tumor-induced pressure.
Third, GlimS developed a method for characterizing the infiltrative and displacive growth tendencies of individual tumors. This approach relies on estimating patient-specific parameters of the mechanically-coupled tumor growth model from clinical Magnetic Resonance (MR) images. After confirming that model parameters are recoverable in an in-silico study, GlimS demonstrated that patient-specific tumor characteristics can be estimated from clinical MR imaging data.
GlimS research was presented at 9 international conferences, 8 national and institutional meetings, and resulted in 3 conference papers and publications so far. Currently, we seek to exploit the results in a follow-up study that focuses on evaluating the clinical predictiveness of model-derived biomechanical tumor characteristics.
