Application of SVD-Based Metabolite Quantification Methods in Magnetic Resonance Spectroscopic Imaging

  • Min Huang
  • Songtao Lu
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 4091)


MRSI can reflect the abnormal metabolites information of different diseases in clinical diagnosis. We made research on the application of SVD-based metabolite quantification methods in 2D MRSI by comparing two different SVD algorithms. In the quantification process, first, the FID signals are rearranged into a data matrix. Then, we can make full SVD by Golub algorithm or partial SVD by Lanczos algorithm. Last, the parameter estimation on each metabolite can be acquired by the definition of the linear parameter model. The ordinary full SVD must decompose all the singular value, with a big cost of the time. The partial SVD just needs to calculate the less singular by the character of the Hankel matrix to improve the estimation speed. When the SNR of MRS signals is higher than ten, the computation time on partial SVD is decreased by thirteen times of the ordinary method. But the speed of quantification is only half of the ordinary one when the SNR is lower than one. Improvements of speed and accuracy in metabolite quantification are key factors for 2D MRSI to be a clinical tool in the future.


Singular Value Decomposition Temporal Lobe Epilepsy Hankel Matrix Magnetic Resonance Spectroscopic Imaging Lanczos Algorithm 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Gruber, S., Stadlbauer, A., Mlynarik, V.: Proton magnetic resonance spectroscopic imaging in brain tumor diagnosis. Neurosurg Clin. N Am. 16, 101–114 (2005)CrossRefGoogle Scholar
  2. 2.
    Vermathen, P., Laxer, K., Schuff, N., Matson, G.: Evidence of neuronal injury outside the medial temporal lobe in temporal lobe epilepsy: N-acetylaspartate concentration reductions detected with multisection proton MR spectroscopic imaging. Radiology 226, 195–202 (2003)CrossRefGoogle Scholar
  3. 3.
    Golub, G., Pereyra, V.: The differentiation of pseudo-inverses and nonlinear least squares problems whose variables separate. SIAM J. Numer. Anal. 10, 413–432 (1973)MATHCrossRefMathSciNetGoogle Scholar
  4. 4.
    Brown, T., Kincaid, B., Ugurbil, K.: NMR chemical shift imaging in three dimensions. Proceedings National Academy of Sciences 79, 3523–3526 (1982)CrossRefGoogle Scholar
  5. 5.
    Vanhamme, L., Sundin, T., Hecke, P.: MR spectroscopy quantitation: a reviews of time-domain methods. NMR in Biomedcine 14, 233–246 (2001)CrossRefGoogle Scholar
  6. 6.
    Golub, G., Reinsch, C.: Singular value decomposition and least squares solutions. Numer. Math. 14, 403–420 (1970)MATHCrossRefMathSciNetGoogle Scholar
  7. 7.
    Laudadio, T., Mastronardi, N., Vanhamme, L., Hecke, P., Huffel, S.: Improved Lanczos Algorithms for Blackbox MRS Data Quantitation. Journal of Magnetic Resonance 157, 292–297 (2002)CrossRefGoogle Scholar
  8. 8.
    Ricardo, D., Eric, P.: Lanczos and the Riemannian SVD in information retrieval applications. Numerical Linear Algebra with applications 3, 1–18 (2003)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2006

Authors and Affiliations

  • Min Huang
    • 1
    • 2
  • Songtao Lu
    • 2
  1. 1.School of Electronic EngineeringSouth-Central University for NationalitiesWuhanChina
  2. 2.Institute of Biomedical EngineeringHuazhong University of Science and TechnologyWuhanChina

Personalised recommendations