基于球面小波的3D曲面形状描述方法及AD脑皮层分类研究

项目名称： 基于球面小波的3D曲面形状描述方法及AD脑皮层分类研究

项目编号： No.61305041

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 闫允一

作者单位： 西安电子科技大学

项目金额： 26万元

中文摘要： 随着数据获取手段的不断发展，三维曲面的应用日益广泛，针对包括大脑皮层表面在内的三维曲面的形状描述研究已经成为热点。虽然有若干的特征适用于三维曲面的检索，但若要对复杂曲面进行形状描述、形状变异检测及分类则尚有许多问题值得研究。申请者在前期研究中发现某些形状特征对曲面形状精细变异有良好的区分度。本项目拟设计新的具备局部化能力和旋转不变性的曲面形状描述符，利用过完备球面小波变换的多尺度特性，对三维曲面进行多尺度形状描述；以新形状描述符为基础，在多尺度空间内，建立阿尔兹海默病(AD)、轻度认知障碍(MCI)和正常人(NC)脑皮层的形状变异模型，揭示变异区域的位置、尺度和幅度等信息；以形状信息为主要依据，设计可对AD、MCI和NC分类的分类器及优化算法，提高分类准确率。本项目成果能用于包括脑皮层在内的三维曲面的描述、形变检测和分类等领域，对推动相关领域的发展起有一定作用，还能为AD分类提供新方法。

中文关键词： 脑皮层曲面；阿尔兹海默症；球面小波；离散粒子群优化；多目标优化

英文摘要： As the development of tools and methods for 3D data gathering, shape feature extraction and reprsentation have oome to being hotspot. Although there are some features suitable for 3D suface search, studies on fine shape detection and classification of complexible surfaces are not well reported. The applicant have found that localized shape feature can be used to screen surfaces' shape variation. This project is planning to design noval shape descriptor with localization ability and multi-scale feature,take the coefficients of over-complete spherical wavelet as the base to describe surface shape in multiscale space; and then this project applys the findings to build a model for cortical surface shape variation betwenn AD,MCI and NC,cotainning information on position, scale and amplitude; and as the last task, the project also tries to design classifier based on the SVM and propose noval optimization algorithm for training.The results of this project can be applied to describe, search and detect variation for 3D surfaces and models, can help to develop methods in ralated fields, and the project can also provide new methods for AD classification.

英文关键词： Surface of cerebral cortex；Alzheimer's disease；Spherical wavelet；Discrete particle swarm optimi；Multi objective optimization