| The effectiveness of TCM is based on its chemical constituents as the materialbasis. Therefore, it is a fundamental study that laboratory analysis methods use kindsof modern analytical instrument, which can lay foundations for laboratory determinepharmacodynamics and pharmacology of active ingredient or effective parts.Gaschromatography-mass spectrometry(GC-MS) integrates chromatographic separationtechnology, mass spectrometry analysis technology and computer technology, whichcombines high separation efficiency of gas chromatography, high resolution of massspectrometry and computing power of computer. GC-MS is well known as a goodqualitative and quantitation technology for complicated components, and it is widelyused in many research fields such as food, medicine, biochemistry and so on. Atpresent, the research on hardware and its supporting software of GC-MS instrumentslags far behind the developed countries, it is just getting started. In order to guaranteethe quality uniform and stability of TCM product, Process Analytical Techniques(PAT) is needed in the production process, monitoring its status and variation ofcontents.It is needed that developing computer algorithm is used to extract feature ofGC-MS data. The had finding algorithms are all designed for single sample, but itcannot meet the fast and exact analysis for massive data. The previous algorithms didnot filter the unwanted mass chromatograms, and its deconvolution algorithm did notwork based on the idear of pattern recognition, the quantifying quantitative rules forselection quantitation ions of a chemical compound is rare reported. Domesticresearch on feature extraction of GC-MS data has not been seen, especially nointegrated GC-MS feature extraction algorithm has been proposed as yet. However,GC-MS has high sensitivity and good seperation, which is slower compaired withNear infrard spectroscopy (NIRS) technology, besides of GC-MS, the paper alsoresearch the NIRS about how to extract feature not deeply.The main works are as follows:(1) The representative feature extraction algorithms of each period since1960s,such as smoothing, correaction, mass chromatograms alignment, peak extraction,deconvolution, qualitative and quantitation are studied. At the same time, three typesof model for GC-MS data is researched. (2) It proposes the following algorithms: unwanted mass chromatograms firtermethod, deconvolution algorithm based on pattern recognition, three-layer parallelarchitecture of GC-MS data processing, quantitative rules of compound in singlesample and algorithms.(3) The proposed algorithms, unwanted mass chromatograms firter method anddeconvolution algorithm is verified by doing standard samples of GC-MSinstrument.It shows that compound can be found more easily and even the ever cannotfind compound because of noising also can be found after filtering the unwanted masschromatograms, at the same time it saves the computing time. After testing the threecases of problems resolved by the proposed deconvolution algorithm, the getted massspectrum of compounds are very similar to the standard mass spectrum itscorresponding to, which fully satisfies the requirement of qualitative identification,matching with the database of mass spectrum.(4) A two-layer parallel computing platform is set up based on a combination ofMPI and Boost thread, the algorithms proposed above are implemented on theplatform, with a speedup of1.9528on two PCs, and a speedup of3.6935on four PCs,witch proves that the platform is well suited for the massive GC-MS data processing.(5)A kind of feature extraction algorithm for NIRS, Laplacian Eigenmapsdimension reduction, is studied. Then a novel online process monitoring methodbased on NIRS is proposed,using to the column chromatographic separation forTraditional Chinese Medicine. |
PDF Full Text Request