Faculty of Engineering and Natural Sciences
Personalizing Tumor Growth Models in Medical Images
Ender Konukoğlu, INRIA, Sophia Antipolis, France
Abstract: In this talk we address the problem of “personalizing mathematical tumor growth models” for the case of brain gliomas using medical images. We work with the reaction-diffusion type growth models which received attention lately due to their applicability on medical images. In terms of general modeling framework these models have been linked with the medical images. The use of anatomical, diffusion and angiography images in the reaction-diffusion modeling have been studied. However, personalizing these model by adapting them to specific patient cases have not been addressed yet. As a first question, we tackle this problem and propose a method for estimating parameters of the reaction-diffusion model through time series of medical images. In this attempt we propose to use a model for the evolution of the tumor delineation in the images based on the same dynamics as the reaction-diffusion models. This model is derived from the reaction-diffusion models however, it does not require the knowledge of tumor cell density distributions throughout the brain. We analyze the estimation methodology theoretically using synthetic experiments and show promising preliminary results on real cases, specifically on the predictive power of personalized reaction-diffusion model. Following this, we focus on the potential use of such a personalized model for the treatment of gliomas. Particularly, we concentrate on radiotherapy and deal with the problem of outlining the irradiation margins. This problem is crucial in treating gliomas as the tumor is infiltrative and medical images do not detect the whole extent of the invasion in the brain. We propose a formulation based on the reaction-diffusion models for extrapolating the invasion extent of gliomas from the part of the tumor visible in the images. In the last part of the talk we focus on numerical methods for anisotropic Eikonal equations. We propose a novel numerical method for solving these equations accurately and fast.
Short Bio: Ender Konukoglu was born in Istanbul on 1981. He got his B.S. and M.S. from the Electrical and Electronics Engineering Department at Bogazici University. He received his PhD in Computer Science from the Universite de Nice-Sophia Antipolis, working under Prof. Nicholas Ayache in the Asclepios Research Project/INRIA Sophia Antipolis. His research focuses on applying mathematical models to the analysis of pathological data. His clinical concentration is on brain tumors and other brain pathologies while methodologically he is interested in partial differential equations, asymptotic analysis, level set methods and stochastic methods.
April 22, 2009, 13:40,