Ruprecht-Karls-Universitšt Heidelberg

Classification of magnetic resonance spectra


Bjoern H. Menze, Michael Kelm, Fred A. Hamprecht, P. Bachert (DKFZ), M. Lichy (DKFZ), H.-P. Schlemmer (Univ. Tuebingen)

Aim

In vivo nuclear magnetic resonance spectroscopy (MRS) opens a window into certain metabolic processes in living tissue. In its clinical application, it is primarily used for the detection and localization of tumorous changes, e.g. in breast, prostate or brain. Magnetic resonance spectroscopic imaging (MRSI) allows to acquire such spectra on two- or three-dimensional grids at resolutions of millimeters and with hundreds to thousand of spectra.

In our work we aim at implementing probabilistic models in the processing of these spatio-spectral data, providing an easier access to relevant diagnostic information in the clinical application of MRSI.

Methods

MR spectra are characterized by a low signal-to-noise ratio and a highly redundant representation of the intrinsic information (Fig. 1). We rely on regularized methods from chemometry and statistical learning to obtain probabilitstic maps indicating tumorous changes of the tissue (Fig. 2). An automated quality assurance based on pattern recognition methods allows for a robust application of these methods in the clinical environment.


 

    

Fig. 1: Spectral signals of tumorous (top) and healthy tissue.

Fig. 2: Classification of spectral images (MRSI). Labels from clinical follow-up examinations:
T/t - tumor, N - normal-state.

 

Publications


  • Automated estimation of tumor probability in prostate MRSI: Pattern recognition vs. quantification.
    Kelm, B.M.; Menze, B.H.; Zechmann, C.M.; Baudendistel, K.T. & Hamprecht, F.A. 
    Magnetic Resonance in Medicine, 2007(57): 150-159
    (preprint)

  • Spectral imaging and applications
    Carlsohn, M.F.; Menze, B.H.; Kelm, B.M.; Hamprecht, F.A.; Kercek, A.; Leitner, R. & Polder, G. 
    in: Lukac, R. & Plataniotis, K.N. (eds.), Color Image Processing: Methods and Applications, CRC Press, to appear 2006
    (publisher) (preprint)

  • Optimal classification of long echo time in vivo magnetic resonance spectra in the detection of recurrent brain tumors
    Menze, B.H.; Lichy, M.P.; Bachert, P.; Kelm, B.M.; Schlemmer, H.-P. & Hamprecht, F.A.
    NMR in Biomedicine, 2006(19): 599 - 609
    (publisher) (preprint)

  • Machine-based rejection of low quality spectra and estimation of brain tumor probabilities from magnetic resonance spectroscopic images.
    Menze, B.H.; Kelm, B.M.;Heck, D.; Lichy, M.P. & Hamprecht, F.A.  
    in: Handels, H. et al. (eds.); Bildverarbeitung für die Medizin 2006, Springer, 2006, 31-36
    (publisher) (preprint)

  • CLARET: a tool for fully automated evaluation of MRSI with pattern recognition methods.
    Kelm, B.M.; Menze, B.H.; Neff, T.; Zechmann, C.M. & Hamprecht, F.A. 
    in: Handels, H. et al. (eds.); Bildverarbeitung für die Medizin 2006, Springer, 2006, 51-55
    (publisher) (preprint)

  • Automated separation of low quality and artifact spectra by pattern recognition in the processing of MR spectral images
    Menze, B.H.; Kelm, B.M. & Hamprecht, F.A.
    Proceedings of the 14th ISMRM Scientific Meeting and Exhibition
    , #2526, Seattle/USA, 2006
    (publisher)

  • [Application of 1H MR Spectroscopic Imaging in Radiation Oncology: Choline as a Marker for Determining the Relative Probability of Tumor Progression after Radiation of Glial Brain Tumors]
    Lichy, M.P. Bachert, P.; Hamprecht, F.A.; Weber, M.-A.; Debus, J.; Schulz-Ertner, D.; Schlemmer, H.-P. & H.-U Kauczor, H.-U.
    Fortschr Rontgenstr 2006(178):627-633, in German
    (publisher)

  • Automatische Lokalisation von Tumoren in 1H-NMR-spektroskopischen in vivo Aufnahmen.
    Kelm, M.; Menze, B. & Hamprecht, F.
    GMA Congress, VDI-Berichte 1883, 2005, 457-466
    (preprint)

  • Classification of in vivo magnetic resonance spectra
    Menze, B.H.; Wormit, M.; Bachert, P.; Lichy, M.P.; Schlemmer, H.-P. & Hamprecht, F.A. 
    in: Weihs, C. & Gaul, W. (eds.), Classification, the ubiquitous challenge. Studies in Classification, Data Analysis, and Knowledge Organization, Vol. 31, Springer, 2005, 362-369
    (publisher) (preprint)

  • Optimal processing in the automatic detection and localization of brain tumors using MRSI.
    Menze, B.H.; Kelm, B.M.; Lichy, M.P.; Bachert, P.; Schlemmer, H. & Hamprecht, F.A.
    Proceedings of the 13th ISMRM Scientific Meeting and Exhibition
    , #2093, Miami/USA, 2005
    (publisher) (preprint)

  • Automated classification of magnetic resonance spectra for the detection of brain tumor
    Menze, B.H.; Wormit, M.; Bachert, P.; Lichy, M.P.; Schlemmer, H.-P. & Hamprecht, F.A.
    in: Detschew, V. et al. (eds.), Biomedizinische Technik, vol. 49, 2004, 438-439

  • Classification of magnetic resonance spectra by means of pattern recognition
    Menze, B.H.; Wormit, M.; Lichy, M.P.; Bachert, P.; Schlemmer, H.-P. & Hamprecht, F.A.
    Proceedings of the 12th ISMRM Scientific Meeting and Exhibition, #2281, Kyoto/Japan, 2004
    (publisher)

 

eXTReMe Tracker


Last update: 06.10.2010, 12:27
zum Seitenanfang