Least squares for diffusion tensor estimation revisited: Propagation of uncertainty with Rician and non-Rician signals

► We derive expressions for the statistics of log-Rician and log-non-central-Chi data. ► We compute the bias and variance in the components of the DT for these signals. ► The bias in the estimation cannot be neglected unless a high SNR is considered. ► The bias is more important in the non-central-Chi case. ► The bias cannot be removed increasing the number of acquired gradient directions.