Error Propagation Framework for Diffusion Tensor Imaging

INTRODUCTION In a typical Diffusion Tensor Imaging (DTI) experiment, generally only one estimate of a tensor is obtained in each voxel. Since the tensor estimate itself is derived from the noisy diffusion-weighted (DW) signals, here we derive the SD of the tensor and tensor-derived quantities by error propagation from the DW signals. This error propagation technique relies on the nonlinear least squares (NLS) objective function of DTI. This proposed technique is shown to produce precise estimate of the SD of FA. The simulation results show that the variability in tensor-derived quantities is largely due to the variability in the reference signal if the DTI model includes the reference signal as a parameter to be estimated. A simple procedure is provided to ameliorate this problem. METHODS Let f = n 6 1 (s − αExp( X β )) 2