Wykorzystanie wielowymiarowych technik analizy widm 1H MRS in vivo w różnicowaniu wrodzonych wad metabolizmu

Abstract

The thesis is based on a series of publications A.1 - A.3. The primary aim was to assess the utility of 1H MRS in vivo technique and multivariate data analysis methods in diagnosis of inborn errors of metabolism in children. 1H MRS in vivo technique provides valuable insight into brain biochemistry. However, relatively low spectral resolution and sensitivity of the technique hamper differentiation of various disorders. Application of multivariate data analysis techniques making use of covariances or correlations between metabolites is expected to facilitate extraction of clinically useful spectral features. In publications A.1 and A.2 visualization of metabolic differences between rare inborn errors of metabolism and other neurological disorders commonly encountered in clinical practice was achieved using dimensionality reduction of 1H MRS data acquired from the regions of interest located in brain white matter. Both water scaled metabolite levels determined with LCModel software (publication A.1) and unresolved 1H MRS in vivo spectra normalized to the sum of low-molecular metabolites (publication A.2) were subjected to principal component analysis. The obtained results proved the usefulness of the method in differentiation of various white matter neurometabolic disiorders (van der Knaap disease, metachromatic leukodystrophy, globoidal leukodystrophy and Canavan disease) from other neurological disorders (cerebral palsy, global developmental delay and epileptic encephalopathy). As neurometabolic disorders individually are extremely rare, it was necessary to pool 1H MRS in vivo data acquired during a relatively long period of time. The publication A.3 was devoted to assessment of long-term MRI scanner reproducibility and application of unsupervised change point detection technique in the analysis of phantom metabolite levels time series. Although multivariate techniques of data analysis are increasingly used for detection of complex 1H MRS derived metabolic signatures in pathological conditions, multivariate analysis of regional heterogeneity of the normal human brain has not been paid attention so far. Thus, the secondary aim of this work was to determine metabolic coordinates of various brain regions in 1H MRS in vivo derived multivariate space. A.1 Polnik A (Skorupa A), Sokół M, Jamroz E, Paprocka J, Wicher M, Banasik T, Marszał E, Kiełtyka A, Konopka M. Contribution of 1H MRS to differential diagnosis of neurologic disorders in children. W: Some aspects of medical physics - in vivo and in vitro studies. Eds.: Z. Drzazga, K. Ślosarek, Polish Journal of Environmental Studies. Series of Monographs. Vol.1, 2010, str. 27-33. A.2. Skorupa A., Jamroz E, Paprocka J, Sokół M, Wicher M, Kiełtyka A. Bridging the gap between metabolic profile determination and visualization in neurometabolic disorders: a multivariate analysis of proton magnetic resonance in vivo spectra. J Chemometrics. 2013;27:76–90. A.3 Skorupa A, Wicher M, Banasik T, Jamroz E, Paprocka J, Kiełtyka A, Sokół M, Konopka M. Fourand- one-half years' experience in monitoring of reproducibility of an MR spectroscopy systemapplication of in vitro results to interpretation of in vivo data. J Appl Clin Med Phys. 2014;15(3):4754.