Comprehensive study of methods for automatic choice of regularization parameter for diffuse optical tomography

Abstract. The image reconstruction in diffuse optical tomography (DOT) is a typical inverse problem; therefore, regularization techniques are essential to obtain a reliable solution. The most general form of regularization is Tikhonov regularization. With any Tikhonov regularized reconstruction algorithm, one of the crucial issues is the selection of the regularization parameter that controls the trade-off between the regularized solution and fidelity to the given sets of data. Automatic methods such as L-curve, generalized cross-validation, minimal residual method, projection error method, and model function method have been introduced to select the regularization parameter over the years. However, little investigation of comparison of all the algorithms has been reported in DOT. The performance of the five methods for choosing regularization parameter is comprehensively compared, and advantages and limitations are discussed.

[1]  S. Arridge,et al.  Optical tomography: forward and inverse problems , 2009, 0907.2586.

[2]  Jaya Prakash,et al.  A LSQR-type method provides a computationally efficient automated optimal choice of regularization parameter in diffuse optical tomography. , 2013, Medical physics.

[3]  Brian W. Pogue,et al.  MR-Guided Near-Infrared Spectral Tomography Increases Diagnostic Performance of Breast MRI , 2015, Clinical Cancer Research.

[4]  Jorge Ripoll,et al.  Feasibility of U-curve method to select the regularization parameter for fluorescence diffuse optical tomography in phantom and small animal studies. , 2011, Optics express.

[5]  K. Kunisch,et al.  Iterative choices of regularization parameters in linear inverse problems , 1998 .

[6]  Yan Zhao,et al.  Direct regularization from co-registered anatomical images for MRI-guided near-infrared spectral tomographic image reconstruction. , 2015, Biomedical optics express.

[7]  Hamid Dehghani,et al.  Improving the quantitative accuracy of cerebral oxygen saturation in monitoring the injured brain using atlas based Near Infrared Spectroscopy models , 2016, Journal of biophotonics.

[8]  G. Golub,et al.  Generalized cross-validation for large scale problems , 1997 .

[9]  Phaneendra K Yalavarthy,et al.  Minimal residual method provides optimal regularization parameter for diffuse optical tomography , 2012, Journal of biomedical optics.

[10]  Eric L. Miller,et al.  Hyperspectral image reconstruction for diffuse optical tomography , 2011, Biomedical optics express.

[11]  S R Arridge,et al.  Recent advances in diffuse optical imaging , 2005, Physics in medicine and biology.

[12]  R. J. Gaudette,et al.  A comparison study of linear reconstruction techniques for diffuse optical tomographic imaging of absorption coefficient. , 2000, Physics in medicine and biology.

[13]  Alexander B. Konovalov,et al.  Space-Varying Iterative Restoration of Diffuse Optical Tomograms Reconstructed by the Photon Average Trajectories Method , 2007, EURASIP J. Adv. Signal Process..

[14]  Brian W. Pogue,et al.  Model-Resolution-Based Basis Pursuit Deconvolution Improves Diffuse Optical Tomographic Imaging , 2014, IEEE Transactions on Medical Imaging.

[15]  Farras Abdelnour,et al.  Hierarchical Bayesian regularization of reconstructions for diffuse optical tomography using multiple priors , 2010, Biomedical optics express.

[16]  E. Miller,et al.  Tomographic optical breast imaging guided by three-dimensional mammography. , 2003, Applied optics.

[17]  Britton Chance,et al.  Diffuse optical tomography with a priori anatomical information , 2003, SPIE BiOS.

[18]  B. Pogue,et al.  Weight-matrix structured regularization provides optimal generalized least-squares estimate in diffuse optical tomography. , 2007, Medical physics.

[19]  Martin Schweiger,et al.  Selection of regularization parameter for optical topography. , 2009, Journal of biomedical optics.

[20]  Yan Zhao,et al.  Optimization of image reconstruction for magnetic resonance imaging–guided near-infrared diffuse optical spectroscopy in breast , 2015, Journal of biomedical optics.

[21]  Tianzi Jiang,et al.  Improving image quality of diffuse optical tomography with a projection-error-based adaptive regularization method. , 2008, Optics express.

[22]  Per Christian Hansen,et al.  Analysis of Discrete Ill-Posed Problems by Means of the L-Curve , 1992, SIAM Rev..

[23]  Kiwoon Kwon,et al.  Uniqueness, Born Approximation, and Numerical Methods for Diffuse Optical Tomography , 2013, J. Appl. Math..

[24]  Yukio Yamada,et al.  Improvement of image quality of time-domain diffuse optical tomography with lp sparsity regularization , 2011, Biomedical optics express.

[25]  S. Arridge.  Optical tomography in medical imaging , 1999 .

[26]  E. Miller,et al.  Quantitative spectroscopic diffuse optical tomography of the breast guided by imperfect a priori structural information , 2005, Physics in medicine and biology.

[27]  Jong Chul Ye,et al.  Compressive Diffuse Optical Tomography: Noniterative Exact Reconstruction Using Joint Sparsity , 2011, IEEE Transactions on Medical Imaging.

[28]  Junzhou Huang,et al.  Diffuse Optical Tomography Enhanced by Clustered Sparsity for Functional Brain Imaging , 2014, IEEE Transactions on Medical Imaging.

[29]  Steven A. Arcone,et al.  Forward and inverse problems in GPR research , 2013 .

[30]  Jun Zou,et al.  An improved model function method for choosing regularization parameters in linear inverse problems , 2002 .