LMS Estimator in Diffusion Tensor Estimation

Abstract:

DWIs do not contain signal bias caused by thermal noise but also outliers introduced by physiological noise. The least square estimator , the classic diffusion tensor estimator in the estimation of DTI, will produce the optimal result in diffusion tensor on the condition that the distribution of the error is Gaussian, while it cannot lead to a robust result if outliers exist. According to the problem exist in LS method, LMS robust estimator is proposed to estimate the diffusion tensor in this paper. Through experiments on the synthetic and real data, and comparing the estimate result of LMS estimator with LS estimator and M estimator, a more robust and more effective result can be produced in diffusion tensor estimation.

• Periodical: Applied Mechanics and Materials (Volume 339)
• Pages: 446-450
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.339.446
• Online since: July 2013
• Authors: San Li Yi, Zheng Gang Liu, Jina Feng He, Lei Ma, Yan Xiang
• Keywords: Diffusion Tensor Image, Least Square Estimator, LMS Robust Estimator, M Robust Estimator

References

[1] P.J. Basser, J.Mattiello and D.LeBihan: Journal of Magnetic Resonance. Series B 103 (1994), p.247

[2] E.O. Stejskal, J.E. Tanner: Journal of Chemical Physics. Vol. 42 (1965), p.288.

[3] Cheng Guan, Lin-Chin Chang: Journal of Magnetic Resonance. Vol.182 (2006), p:115

[4] J-F Mangin, C. Poupon, C. Clark: Medical Image Analysis. Vol.6 (2002), p.191

[5] Lin-Ching Chang, Derek K Jones: Magnetic Resonance in Medicine. Vol.53(2005), p.1088

[6] P.J. Rousseeuw: J Am Stat Assoc.Vol.79(1984),p.871

[7] R.L. Plackett: Biometrika. Vol.37(1950), p.147

[8] Victor J. Yohai: Journal of American Statistical Association. Vol.83(1988), p.406