Accounting for Anisotropic Growth of Glioma in Radiotherapy Planning
J Unkelbach1*, B Menze2, E Konukoglu3, A Motamedi4, N Ayache5, H Shih6, (1) Massachusetts General Hospital, Boston, MA, (2) ETH Zurich, Switzerland, (3) Microsoft Research, Cambridge, UK, (4) Massachusetts General Hospital, Boston, MA, (5) INRIA Sophia Antipolis, France (6) Massachusetts General Hospital, Boston, MASU-C-BRB-5 Sunday 1:30:00 PM - 2:15:00 PM Room: Ballroom B
Purpose: We study a phenomenological tumor growth model for improved target volume definition for radiotherapy of glioblastoma. Currently, an isotropic margin is added to the visible tumor to account for microscopic tumor cell infiltration in normal appearing brain tissue. However, it is known that glioma growth is not isotropic. The ventricular system and the dura, including falx cerebri and tentorium cerebelli, represent anatomical barriers for migrating cells. Such anatomical constraints are currently not consistently and quantitatively incorporated in target delineation.
Methods: We assume that tumor growth is characterized by local proliferation of tumor cells and diffusion into neighboring tissue. Mathematically, this can be described via a partial differential equation of reaction-diffusion type, the Fisher-Kolmogorov-Equation. Anatomical constraints are modeled via no-flux boundary. Solving the model equations provides a three-dimensional distribution of the tumor cell density. The radiotherapy target can be defined as an iso-line of the cell density.
Results: Two specific questions were investigated: First, tumor locations in which the model may be particularly useful have been identified; and second, a sensitivity analysis with respect to the model inputs was performed. Tumors located in proximity to the falx cerebri may benefit: The model is able to describe both the falx as a boundary and the corpus callosum, which provides a route for tumor cells to spread to the contralateral hemisphere. This effect is often not consistently accounted for in manual target delineation. Among all model parameters, correct segmentation of the brain is a critical prerequisite.
Conclusions: The tumor growth model represents a tool to objectively create target volumes for radiotherapy of glioblastoma by consistently accounting for known growth patterns. All model inputs can be linked to model outputs, making it possible to assess the impact of uncertainties - an essential step for an application in clinical practice.