Novel Approaches Of Electrogastrography Detection And Analysis

Cutaneous Electrogastrography(EGG) is one of a important tools for GI motility studies. However, current measurement systems and analysis methods for clinic applications still remain at the level of 1980's and 1990's and need to be updated. Therefore, a novel solution including sampling, prepocessing, analysis is presented in this dissertation.The main research content of the dissertation involves:1) A mathematical model of slow wave is developed, and the background noise is simulated in order to supervise the hardware design and evaluate the performance of preprocessing algorithm.2) The feasibility of introducing Body Surface Laplacian technique into EGG exploration is discussed. Synthesizing theoretical analysis and engineering practice, some detailed rules about the parameter optimization of Laplacian electrodes are presented. Consequently, as a instance, a setof EGG Laplacian electrodes is designed.3) Considering the features of EGG, which has ultra-weak amplitude, super-low frequency and extremely poor SNR, the basic principle of hardware design, which <疤?charactered by small gain and wide dynamic range is proposed. Two systems are established respectively, o??f which is kerneled with a high resolution 24 bitΣ-ΔADC and aimed at EGG collecting. The other one is designed as an universal bioelectricity detector. Taking advantages of oversampling technique, the resolution of latter can reach the same standard as the former, simultaneously fill up the deficiency of low speed ofΣ-ΔADCs. Furthermore, since the sampling rate and efficient resolution of the system is adjustable, the system is suitable for all-purpose utilization. Additionally, in this section, a method of getting mucous EGG by using electronic capsule is presented.4) EGG is a typical non-linear and non-stationary signal. And due to individual disparities, all kind of experimental knowledge is quite difficult to acquire. Accordingly, during the stage of signal preprocessing, blind and adaptive methods are preferentially considered. Empirical Mode Decomposition(EMD) is introduced to filtrate the interference of ECG and respireatory. For multi-channel signals, Independent Component Analysis(ICA) is employed to reject noises synchronously. An algorithm based on difference replacement is prompted to remove gross errors, which are caused by motional artifical and baseline drift. The method is proved to be efficient.5) As an utility of EMD algorithm, Hilbert spectrum and Hilbert marginal spectrum is calculated to take place of conventional power spectrum. By this method, resolutions both in time and frequency domain can be obviously improved.6) Normal and abnormal EGG data recorded from 38 subjects were classified and analyzed using Support Vector Machine(SVM). And experiment results showed that the discrimination and the generalization ability of SVM is higher than traditional ANN methods. Finally, advices about the modification of current EGG diagnostic standard is prompted.