Magnetic Resonance Image Segmentation with Thin Plate Spline Thresholding

We propose a new method for the T1-weighted magnetic resonance image (MRI) segmentation. Thin plate splines are fitted to overlapping blocks of an image slice and thresholds are found. The knots and the smoothing parameters of the splines are chosen by a modified version of the generalized cross validation criterion. Each block is associated with a weighting function, which serves to blend the splines together as well as the thresholds in a smooth fashion. The blended image is then thresholded to get the boundaries between gray matter, white matter, cerebrospinal fluid, and others. We tested the method on MGH CMA 20 normal data. The results show that our method achieves good segmentation compared to human segmentation and SPM segmentation. Also our method generates subpixel results and handles the partial volume effect in the model. The new method has the advantage of being less dependent on image non-uniformity correction.

[1]  David A. Rottenberg,et al.  Quantitative comparison of four brain extraction algorithms , 2004, NeuroImage.

[2]  G. Wahba.  Spline models for observational data , 1990 .

[3]  Yianni Attikiouzel,et al.  A probabilistic neural network based image segmentation network for magnetic resonance images , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[4]  James A. Sethian,et al.  Level Set Methods and Fast Marching Methods: Evolving Interfaces in Computational Geometry, Fluid , 2012 .

[5]  Alan C. Evans,et al.  Fast and robust parameter estimation for statistical partial volume models in brain MRI , 2004, NeuroImage.

[6]  G. Wahba,et al.  Gcvpack – routines for generalized cross validation , 1987 .

[7]  L O Hall,et al.  Review of MR image segmentation techniques using pattern recognition. , 1993, Medical physics.

[8]  Guang H. Yue,et al.  Automated Histogram-Based Brain Segmentation in T1-Weighted Three-Dimensional Magnetic Resonance Head Images , 2002, NeuroImage.

[9]  Sun-Yuan Kung,et al.  Quantification and segmentation of brain tissues from MR images: a probabilistic neural network approach , 1998, IEEE Trans. Image Process..

[10]  Moo K. Chung,et al.  Less white matter concentration in autism: 2D voxel-based morphometry , 2004, NeuroImage.

[11]  Benoit M. Dawant,et al.  Morphometric analysis of white matter lesions in MR images: method and validation , 1994, IEEE Trans. Medical Imaging.

[12]  Karl J. Friston,et al.  Voxel-based morphometry , 2007 .

[13]  Anders M. Dale,et al.  Cortical Surface-Based Analysis I. Segmentation and Surface Reconstruction , 1999, NeuroImage.

[14]  Benoit M. Dawant,et al.  Neural-network-based segmentation of multi-modal medical images: a comparative and prospective study , 1993, IEEE Trans. Medical Imaging.

[15]  Karl J. Friston,et al.  Incorporating Prior Knowledge into Image Registration , 1997, NeuroImage.

[16]  Moo K. Chung,et al.  Subpixel Curvature Estimation of the Corpus Callosum via Splines and its Application to Autism , 2004 .

[17]  C. Davatzikos,et al.  Using a deformable surface model to obtain a shape representation of the cortex , 1995, Proceedings of International Symposium on Computer Vision - ISCV.

[18]  Natasa Kovacevic,et al.  A Robust Method for Extraction and Automatic Segmentation of Brain Images , 2002, NeuroImage.

[19]  Sally Wood,et al.  Bayesian mixture of splines for spatially adaptive nonparametric regression , 2002 .

[20]  G. Wahba,et al.  Hybrid Adaptive Splines , 1997 .

[21]  W. Eric L. Grimson,et al.  Segmentation of brain tissue from magnetic resonance images , 1995, Medical Image Anal..

[22]  S. Osher,et al.  Geometric Level Set Methods in Imaging, Vision, and Graphics , 2011, Springer New York.

[23]  J. Ashburner,et al.  Multimodal Image Coregistration and Partitioning—A Unified Framework , 1997, NeuroImage.

[24]  Karl J. Friston,et al.  Voxel-Based Morphometry—The Methods , 2000, NeuroImage.

[25]  Alan C. Evans,et al.  Automated 3-D Extraction of Inner and Outer Surfaces of Cerebral Cortex from MRI , 2000, NeuroImage.