We present the SIBIA framework for joint image registration and biophysical inversion. Given an atlas brain MRI and a cancer patient MRI, we invert for patient-specific tumor growth parameters. We couple diffeomorphic image registration between atlas and patient with a reaction-diffusion tumor model. We derive Picard iterative solver schemes for the PDE-constrained optimization formulation of the coupled problem. Several tests for synthetic and clinical datasets assess the performance and convergence in different usage scenarios.

This presentation is part of Minisymposium “MS30 - Imaging, Modeling, Visualization and Biomedical Computing (2 parts)”
organized by: Cristian Linte (Biomedical Engineering and Center for Imaging Science, Rochester Institute of Technology) , Suzanne Shontz (University of Kansas) .

Authors:

Klaudius Scheufele (University of Stuttgart)

Andreas Mang (Department of Mathematics, University of Houston)

Amir Gholami (UC Berkeley)

Christos Davatzikos (University of Pennsylvania)

Miriam Mehl (Universitat Stuttgart)

George Biros (Institute for Computational Engineering and Sciences, University of Texas at Austin)

Keywords:

brain tumor, high-performance computing, image registration, inverse problems, nonlinear optimization, partial differential equation models