The purpose of this work was two-fold. In the first instance, 1H NMR spectra of the ultracentrifuged lipoprotein fractions (VLDL, LDL and HDL) from six volunteers with different clinical conditions were measured. The methylene regions of the experimental spectra were modelled in the frequency domain using non-linear lineshape fitting analyses. In this way the resolvable Lorentzian component structures of the methylene regions of these lipoprotein fraction spectra could be determined. Second, the lipoprotein fraction analyses were used to construct simplified component structures, which interpreted the lipoprotein fraction spectra well, and were feasible to use in the total plasma spectra analyses. The considerable overlap problem of the resonances was properly handled in this way. The NMR-based relative amounts of the lipoproteins (relative integrated intensities of the lipoprotein model signals) obtained were compared to the biochemically resolved relative molar percentages of the lipoprotein fractions and also of the lipid contents between the lipoprotein complexes. It was noticed that nearly all correlations were extremely good. Thus, it is suggested that the developed methodology could be used as a fast method to predict the relative amounts of the lipoproteins and also possibly the relative lipid contents between the major lipoprotein categories directly from the proton NMR spectrum of a total blood plasma sample. Furthermore, if internal or external reference for the integrated intensities of the proton NMR resonances were used, it should also be possible to obtain the absolute amounts of these quantities.
"GP gaussian processes GPCh glycerphosphocholine GPLS generalized partial least squares GSH glutathione HLSVD Hankel Lanczos singular value decomposition  HLSVD-IRL HLSVD with implicitly restarted Lanczos algorithm  HLSVD-PRO HLSVD with partial reorthogonalization  HR-MAS high resolution magic angle spinning HSVD Hankel singular value decomposition  HTLS Hankel total least squares  HTLS-PK Hankel total least squares using prior knowledge  ICA independent component analysis IQML iterative quadratic maximum likelihood  KLR kernel logistic regression  KNOB-TLS knowledge based total least squares  Lac lactate LCModel linear combination of model spectra  LDA linear discriminant analysis LE long echo time LF lineshape fitting  LGA low grade astrocytoma Lip1 lipids at 1.3 ppm Lip2 lipids at 0.9 ppm LP linear prediction LS least-squares LS-SVM least-squares support vector machines MeFreS Metropolis frequency-selective  MEN meningioma MET metastasis MM macromolecule MODE method of direction estimation  MP matrix pencil  MP-FIR maximum-phase FIR  MR magnetic resonance MRS magnetic resonance spectroscopy MRSI magnetic resonance spectroscopic imaging Myo myo-inositol NAA N-acetyl-aspartate NLLS nonlinear least-squares NMR nuclear magnetic resonance NMR-SCOPE NMR spectra calculation using operators  PCA principal component analysis PCh phosphocholine PET Positron Emission Tomography PLS partial least squares PM performance measure xi PR pattern recognition PSR P-spline signal regression QUALITY quantification improvement by converting lineshapes to the Lorentzian type  QUECC QU from QUALITY and ECC  QUEST quantitation based on quantum estimation  RBF radial basis function RF random forest RRMSE relative root mean squared error SB-HOYWSVD sub-band high-order Yule-Walker singular value decomposition  SE short echo time SELF-MODE selective-frequency MODE  SELF-SVD selective-frequency singular value decomposition  SNR signal-to-noise ratio SVD singular value decomposition SVM support vector machines Tau taurine tCho total choline tCr total creatine TDFD time-domain frequency-domain TLS total least squares VARPRO variable projection  VOI volume of interest "
[Show abstract][Hide abstract] ABSTRACT: The topic of this thesis belongs to the wide field of biomedical signal processing. The main objective is to improve brain tumor diagnosis based on Magnetic Resonance Spectroscopic (MRS) and Imaging (MRI) data. MRI data are 2D grayscale images fully characterized by the intensity of the pixels. MRS data are signals, usually modeled in the time domain as a sum of decaying complex exponentials. The methods developed or used in this thesis are signal processing or pattern recognition methods. Signal processing methods aimed to clean the data (denoising, removing artefacts, etc) and quantify the metabolite concentrations. For cleaning the data, we use a variety of methods ( FIR filter, state-space, wavelets, smoothing techniques ). When quantifying MRS data, we try to minimize the fitting error assuming parametric or semi-parametric models. In this thesis the nonlinear least squares problem is solved with the Levenberg-Marquardt optimization algorithm. The nonparametric part of the model is a sum of penalized splines in the frequency domain, transformed back into the time domain where the optimization is done. Pattern recognition is composed of feature extraction and classification. In this thesis, we tested several feature extraction methods: PCA, ICA, Kruskal-Wallis, Fisher criterion, Relief-F ; and several classification algorithms: LDA, LS-SVM, Random Forests , ... Doctor in de ingenieurswetenschappen Afdeling ESAT - SCD: SISTA/COSIC/DOCARCH Departement Elektrotechniek (ESAT) Faculteit Ingenieurswetenschappen Doctoral thesis Doctoraatsthesis
[Show abstract][Hide abstract] ABSTRACT: A comparison between a time domain analysis algorithm (VARPRO) and a frequency domain analysis algorithm (FITPLAC) for parameter estimation of magnetic resonance spectroscopy (MRS) data series is presented. VARPRO analyses the measured MRS signal (free induction decay; FID); FITPLAC analyses the discrete Fourier transform of the FID, the frequency domain magnetic resonance spectrum. A rapid time domain method, used to subtract the dominating water resonance from a 1H MRS FID, without affecting the metabolites of interest, is outlined and applied. Also a new "pseudofrequency selective" approach to time domain fitting is introduced. The possibilities of combining the most favorable features of time and frequency domain processing into one single MRS signal processing method are assessed. The 1H MRS signals of ultracentrifuged very low (VLDL), intermediate (IDL), and high (HDL) density lipoprotein fractions from human blood plasma were used for the comparisons. The results from both algorithms were in good agreement.
Magnetic Resonance in Medicine 04/1994; 31(4):347-58. DOI:10.1002/mrm.1910310402 · 3.57 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: We demonstrate that information on internal orientational order and size of lipoprotein particles can be extracted from the positions of their NMR spectral lines. The magnetic field obtained by solving the field equations for a model lipoprotein particle is shown to account for the hitherto unexplained size dependence of the experimental NMR frequencies. The predicted sign, magnitude, and functional form of the frequency shifts are verified by novel experimental 1H NMR data from size-specifc lipoprotein samples.
Data provided are for informational purposes only. Although carefully collected, accuracy cannot be guaranteed. The impact factor represents a rough estimation of the journal's impact factor and does not reflect the actual current impact factor. Publisher conditions are provided by RoMEO. Differing provisions from the publisher's actual policy or licence agreement may be applicable.