Weighted linear least squares estimation of diffusion MRI parameters: Strengths, limitations, and pitfalls

PURPOSE Linear least squares estimators are widely used in diffusion MRI for the estimation of diffusion parameters. Although adding proper weights is necessary to increase the precision of these linear estimators, there is no consensus on how to practically define them. In this study, the impact of the commonly used weighting strategies on the accuracy and precision of linear diffusion parameter estimators is evaluated and compared with the nonlinear least squares estimation approach. METHODS Simulation and real data experiments were done to study the performance of the weighted linear least squares estimators with weights defined by (a) the squares of the respective noisy diffusion-weighted signals; and (b) the squares of the predicted signals, which are reconstructed from a previous estimate of the diffusion model parameters. RESULTS The negative effect of weighting strategy (a) on the accuracy of the estimator was surprisingly high. Multi-step weighting strategies yield better performance and, in some cases, even outperformed the nonlinear least squares estimator. CONCLUSION If proper weighting strategies are applied, the weighted linear least squares approach shows high performance characteristics in terms of accuracy/precision and may even be preferred over nonlinear estimation methods.

[1]  D. Noll,et al.  Homodyne detection in magnetic resonance imaging. , 1991, IEEE transactions on medical imaging.

[2]  P. Basser,et al.  The b matrix in diffusion tensor echo‐planar imaging , 1997, Magnetic resonance in medicine.

[3]  Robin M Heidemann,et al.  Generalized autocalibrating partially parallel acquisitions (GRAPPA) , 2002, Magnetic resonance in medicine.

[4]  Carl-Fredrik Westin,et al.  Least squares for diffusion tensor estimation revisited: Propagation of uncertainty with Rician and non-Rician signals , 2012, NeuroImage.

[5]  Jan Sijbers,et al.  Maximum-likelihood estimation of Rician distribution parameters , 1998, IEEE Transactions on Medical Imaging.

[6]  P. Basser,et al.  MR diffusion tensor spectroscopy and imaging. , 1994, Biophysical journal.

[7]  M. Horsfield,et al.  Optimal strategies for measuring diffusion in anisotropic systems by magnetic resonance imaging , 1999, Magnetic resonance in medicine.

[8]  W. Hoge,et al.  Statistical noise analysis in GRAPPA using a parametrized noncentral Chi approximation model , 2011, Magnetic resonance in medicine.

[9]  Anders Kristoffersen,et al.  Estimating non‐gaussian diffusion model parameters in the presence of physiological noise and rician signal bias , 2012, Journal of magnetic resonance imaging : JMRI.

[10]  William E. Lorensen,et al.  The NA-MIC Kit: ITK, VTK, pipelines, grids and 3D slicer as an open platform for the medical image computing community , 2006, 3rd IEEE International Symposium on Biomedical Imaging: Nano to Macro, 2006..

[11]  Jerry L. Prince,et al.  Diffusion Tensor Estimation by Maximizing Rician Likelihood , 2007, 2007 IEEE 11th International Conference on Computer Vision.

[12]  P. Basser,et al.  A unifying theoretical and algorithmic framework for least squares methods of estimation in diffusion tensor imaging. , 2006, Journal of magnetic resonance.

[13]  J Sijbers,et al.  Multiscale white matter fiber tract coregistration: A new feature‐based approach to align diffusion tensor data , 2006, Magnetic resonance in medicine.

[14]  P. Basser,et al.  Toward a quantitative assessment of diffusion anisotropy , 1996, Magnetic resonance in medicine.

[15]  S. Das,et al.  Regional Values of Diffusional Kurtosis Estimates in the Healthy Brain during Normal Aging , 2017, Clinical Neuroradiology.

[16]  Thomas R. Knösche,et al.  White matter integrity, fiber count, and other fallacies: The do's and don'ts of diffusion MRI , 2013, NeuroImage.

[17]  S. Aja‐Fernández,et al.  Influence of noise correlation in multiple‐coil statistical models with sum of squares reconstruction , 2012, Magnetic resonance in medicine.

[18]  J. Sijbers,et al.  Constrained maximum likelihood estimation of the diffusion kurtosis tensor using a Rician noise model , 2011, Magnetic resonance in medicine.

[19]  Alexander Leemans,et al.  The B‐matrix must be rotated when correcting for subject motion in DTI data , 2009, Magnetic resonance in medicine.

[20]  Heidi Johansen-Berg,et al.  Diffusion MRI at 25: Exploring brain tissue structure and function , 2012, NeuroImage.

[21]  Daniel C. Alexander,et al.  Camino: Open-Source Diffusion-MRI Reconstruction and Processing , 2006 .

[22]  M. Johansson,et al.  Regional Values of Diffusional Kurtosis Estimates in the Healthy Brain , 2012, Journal of magnetic resonance imaging : JMRI.

[23]  Jan Sijbers,et al.  ExploreDTI: a graphical toolbox for processing, analyzing, and visualizing diffusion MR data , 2009 .

[24]  Anders Kristoffersen,et al.  Optimal estimation of the diffusion coefficient from non-averaged and averaged noisy magnitude data. , 2007, Journal of magnetic resonance.

[25]  J. Helpern,et al.  Three‐dimensional characterization of non‐gaussian water diffusion in humans using diffusion kurtosis imaging , 2006, NMR in biomedicine.

[26]  Derek K. Jones,et al.  “Squashing peanuts and smashing pumpkins”: How noise distorts diffusion‐weighted MR data , 2004, Magnetic resonance in medicine.

[27]  V. Kiselev.  The Cumulant Expansion: An Overarching Mathematical Framework For Understanding Diffusion NMR , 2010 .

[28]  E. Bullmore,et al.  Formal characterization and extension of the linearized diffusion tensor model , 2005, Human brain mapping.

[29]  P. Basser,et al.  Estimation of the effective self-diffusion tensor from the NMR spin echo. , 1994, Journal of magnetic resonance. Series B.

[30]  J. Helpern,et al.  Diffusional kurtosis imaging: The quantification of non‐gaussian water diffusion by means of magnetic resonance imaging , 2005, Magnetic resonance in medicine.

[31]  A. Leemans,et al.  Comprehensive framework for accurate diffusion MRI parameter estimation , 2013, Magnetic resonance in medicine.

[32]  J. Sijbers,et al.  More accurate estimation of diffusion tensor parameters using diffusion kurtosis imaging , 2011, Magnetic resonance in medicine.

[33]  Cheng Guan Koay,et al.  A signal transformational framework for breaking the noise floor and its applications in MRI. , 2009, Journal of magnetic resonance.

[34]  P. Basser,et al.  Comprehensive approach for correction of motion and distortion in diffusion‐weighted MRI , 2004, Magnetic resonance in medicine.

[35]  Jesper L. R. Andersson,et al.  Maximum a posteriori estimation of diffusion tensor parameters using a Rician noise model: Why, how and but , 2008, NeuroImage.