The Application of EMD to Diffusion Tensor Imaging for Schizophrenia Characterization

Abstract

Schizophrenia is a chronic severe brain illness In recent years connectivity of nerve in white matter becomes an important issue in schizophrenia and DTI have been proved as a powerful technique on white matter Although the voxelwise analysis method can provide a more objective method comparing to the ROI method it would lost the information when do the steps such like registration and smoothing To improve the shortcomings in voxelwise analysis we developed a feature characterization system for image based on empirical mode decomposition (EMD) and filtered back projection In this study we discussed the problems in EMD such like cubic spline boundary effect and mode mixing and proposed the modified method With the proposed system we can obtained the intrinsic mode function (IMF) images respectively which is thought to contain different physical meanings in different IMFs Then we combined with TBSS and observed the IMF enhanced image would result in any effect on this method From results we found that the high frequency components such like IMF1 and IMF2 can enhance the minor tract information and help us find the abnormal region in schizophrenia which is hard to find in traditional method In this study we found that the abnormal regions in fornix left inferior longitudinal fasciculus left superior longitudinal fasciculus which were hard to find in traditional method The low frequency components such like IMF3 and IMF4 enhance the major white matter tract but the statistic results were almost same as traditional method In summary our method can extract the feature and enhance the specific information on the voxelwise analysis which can provide another way to analyze the abnormality area in schizophrenia different from traditional way

Date of Award	2014 Aug 29
Original language	English
Supervisor	Kuo-Sheng Cheng (Supervisor)