Advancing Mathematical Tumor Growth Models for Glioblastoma Multiforme:

Read full text Summarise with VIRAT
Authors: Rajit Chatterjea, Erin J. Youn
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Glioblastoma multiforme (GBM) is the most aggressive primary brain tumor in humans, with a median survival of approximately 15 months. Mathematical tumor growth models based on the Fisher–Kolmogorov reaction-diffusion partial differential equation (PDE) offer principled predictions of occult tumor spread beyond the visible lesion boundary, yet have not entered clinical practice due to computational cost, lackof patient specificity, and restrictive biological assumptions. Jain et al. [1] recently demonstrated that a deep encoder-decoder neural network (GlioMod) can serve as a fast surrogate for the PDE solver, achieving two orders-of-magnitude speed-up while maintaining R2 = 0.95 agreement with ground-truth PDE solutions. This work proposes five mathematically rigorous advancements to the GlioMod framework that address its most critical limitations. First, we extend the scalar isotropic diffusion coefficient to a full diffusion tensor derived from diffusion-tensor MRI (DTI), enabling fiber-tract-guided anisotropic tumor spread. Second, we embed a physics-informed learning objective that enforces PDE residual constraints during training, guaranteeing physical consistency of neural predictions. Third, we intro duce a Bayesian parameter estimation module using amortized variational inference to produce patient-specific, uncertainty-quantified distributions over the diffusivity and proliferation parameters. Fourth, we replace the fixed-resolution encoder-decoder back bone with a Fourier Neural Operator (FNO) that operates directly in function space, eliminating the resolution bottleneck that forces GlioMod to downsample clinical im ages. Fifth, we couple the reaction-diffusion PDE to a linear elastic model of brain deformation, enabling the surrogate to capture mass-effect-driven changes in tumor morphology. Together, these advancements move toward a mathematically complete, clinically deployable GBM modeling framework.

Fields of Study & Subjects

Mathematical Biology Tumor Growth Glioma Diagnosis and Treatment Brain Tumor Detection and Classification

Publication details

Journal / source
Figshare
Publication date
28 March 2026
Language
EN
Document type
Preprint