Drew Morris1, Revital Nossin-Manor1, Margot J. Taylor1, John G. Sled2,3
1Diagnostic Imaging, The Hospital for Sick Children, Toronto, Ontario, Canada; 2Physiology Experimental Medicine, The Hospital for Sick Children, Toronto, Ontario, Canada; 3Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada
DTI was performed on 19 premature neonates (mean post conceptional age 29 weeks). Motion/eddy current correction, B0 field inhomogeneity correction and outlier rejection were applied to the data prior to a nonlinear tensor fit. Goodness of fit in the model was evaluated by measuring the mean sum squared residual before and after each processing step. Each step was found to reduce model residuals, thereby improving the model. Fractional anisotropy was also found to decrease with preprocessing. It was also demonstrated that removing outliers before other preprocessing steps showed better data improvement than doing so afterwards in voxels with higher anisotropy.