VISUALIZING BRAIN FUNCTIONS AND BRAIN VASCULAR DYNAMICS WITH DIFFUSE OPTICAL TOMOGRAPHY: A COMBINED APPROACH FOR THE INVERSE PROBLEM

Abstract

In this thesis, we focused on the improvement of inverse algorithms used in 3D reconstruction for functional brain activation and brain vascular dynamics. A novel inverse algorithm was developed that combines Truncated Singular Value Decomposition, Compressive Sensing, depth compensation and Least Square Method with regularization. This algorithm can achieve higher reconstruction accuracy compared to single reconstruction method in simulations. For example, it reduced the reconstruction error from 50% to 30% for 2D examples and from 60% to 40% for 3D cases when 3% Gaussian white noise was added. A priori structure information was used to improve the image quality with the error reduced to 7% and 15% for 2D and 3D examples, respectively. The algorithm was validated with liquid phantom experiments with good reconstruction accuracy in terms of value restored and depth localization. It was tested with human functional experiment and human brain vascular dynamics experiment with acceptable reconstruction results.