Application Of Empirical Mode Decomposition In Medical MR Image Feature Extraction

Medical image feature extraction is one of the basic content of the medical image processing and analysis. It is a critical step towards computer-aided diagnosis. The aim of medical image feature extraction is in order to obtain medical images’ unique characteristics information that different from other image, and then further achieving computer-aided diagnosis functions such as automatic classification and matching of medical images.The MR images Generated by the magnetic resonance image MRI technology are with low contrast, uneven gray and weak boundary characteristics. Hilbert-Huang Transform is presented by the American Academy of Engineering, the Chinese-American named Huang and it is a new adaptive data analysis methods. Empirical mode decomposition algorithm is used to get a series of intrinsic mode functions and it can be applied to non-stationary and nonlinear data. The research in the field of1-dimensional signal and2-dimensional image signal analysis and processing has become a hot research topic.According to the research status of medical image feature extraction and empirical mode decomposition algorithm, this paper will study the applications of empirical mode decomposition algorithm in medical MR image feature extraction. The main work of this paper is as follows:(1) A serious in-depth study of the basic theories of empirical mode decomposition algorithm. Analyzed the research status in the field of medical image feature extraction and overviewed the current situation of empirical mode decomposition algorithm at home and abroad;(2) Studied the development and application empirical mode decomposition algorithm in the field of two-dimensional images, in-depth studied the extraction algorithm of extreme points in BEMD, surface interpolation algorithm of the envelope, boundary effects and other key problems. Using Delaunay triangle method to implement the interpolation of extremes reduced time consuming and accorded with the properties of medical images. (3) Studied the classical mode aliasing method, ensemble empirical mode decomposition, according to the ideas of noise-assisting, implemented the two-dimensional ensemble empirical mode decomposition algorithm. Comparing study showed the advantage of two-dimensional ensemble empirical mode decomposition.(4) Studied the Riesz transform, then, extracted the local features, such as local amplitude, local phase, based on the Riesz transform and two-dimensional ensemble empirical mode decomposition. Proposed a new edge extraction method that based on the phase congruency and two-dimensional ensemble empirical mode decomposition, clinical data experiment proved its feasibility and validity.