Inverse Scattering Models and Algorithms for Functional Brain Imaging with Diffuse Optical Wavefields

Northeastern University, Boston MA

Investigators

Abstract

0139968 Miller This project is concerned with modeling and algorithmic issues surrounding functional brain imaging from diffuse optical wave data. The primary objective is to efficiently obtain high-quality spatial and temporal three-dimensional image maps of hemodynamic processes in the brain from sparsely sampled scattered wavefield data in conjunction with auxiliary data collected from fMRI scans. The specific aims are: 1) development of spatial and temporal inversion algorithms that exploit the high spatial resolution of simultaneously available MRI and fMRI data as well as the high temporal resolution of diffuse optical wavefield data, 2) fast solution of the underlying forward problem that models that physics of the diffuse optical scattering problem and 3) validation of the algorithms and efficiency using data from experimental studies. One level of efficiency and accuracy will be achieved by developing a low-dimensional, spatially adaptive parameterization of the unknown physical parameters (specifically, optical scattering and absorption coefficients) that reduces the ill-posedness of the inverse problem and concentrates degrees of freedom so as to focus resolution over cortical regions of interest. Further spatial accuracy will be achieved by constraint-based incorporation of anatomical information provided by MRI and fMRI images. To attain better computational efficiency, the proposed work will also focus on parallel, preconditioned Krylov subspace methods for solving the linearized subproblems that arise at every iteration of the inversion process. In the interest of amortizing the cost in temporal imaging, the proposed research will also address how to optimally exploit information obtained from previous reconstructions. All modeling and algorithmic work will be validated and modified based on experimentally obtained diffuse optical and MRI data.