Fig 3.

Impact of diffusion tensor estimation method on tract reconstruction of poor-quality DTI data. Characteristic representations illustrate the effect of tensor estimation methodology on reconstruction of the fornix with high percentage of data outliers (>10%). Note that reconstruction is not possible with the use of the linear least squares (LLS) and weighted linear least-squares (WLLS) methods and appears to be slightly possible with nonlinear least squares (NLLS) but is very well performed if the robust estimation of tensors by outlier rejection (RESTORE) approach is used.