Comparison of Reconstruction Methods for Multi-compartmental Model in Diffusion Tensor Imaging

Diffusion tensor imaging (DTI) is one of the magnetic resonance techniques to describe the anisotropic diffusion in terms of its orientation. DTI gives the direction of white matter fibers in a single direction. However, multi-fiber heterogeneity can be present at several places of the human brain. Recently, a multi-compartmental model (which uses noncentral Wishart distributions) was proposed to improve the state of the art of solving this multi-fiber heterogeneity. In this model, nonnegative least square (NNLS) method was used for solving the inverse problem which is based on L-2 norm minimization. In this paper, results are obtained with the least absolute shrinkage and selection operator (L-1 regularization). In particular, we study the performance of NNLS and nonnegative lasso methods and shown that the later method outperforms for several cases.