基于径向基函数神经网络的两步核磁共振头部组织电阻抗成像算法  

A Two-step MREIT Algorithm for Head Tissues Based on Radial Basic Function Neural Network

作　　者：闫丹丹[1] 张孝通[1] 朱善安[1] Bin He 

机构地区：[1]浙江大学电气工程学院,浙江杭州310027 [2]Department of Biomedical Engineering, University of Minnesota,USA

出　　处：《航天医学与医学工程》2007年第2期126-132,共7页Space Medicine & Medical Engineering

基　　金：国家自然科学基金资助项目(5057705);美国国家科学基金(NSFBES-0411898);美国国立卫生院基金(NIHR01EB00178)

摘　　要：目的基于径向基函数(radial basic function,RBF)神经网络的两步核磁共振电阻抗成像(mag-netic resonance electrical impedance tomography,MREIT)算法,对人体头部进行MREIT。方法首先利用高分辨率的核磁共振成像(magnetic resonance imaging,MRI)系统对成像物体进行三维构建和不同组织的边界区分;然后利用RBF MREIT方法对物体内不同组织的均匀电阻抗分布进行估计,并采用基于径向基函数-遗传算法的MREIT技术对每种组织有限元模型中每个单元的电阻抗值进行估计。结果在三层球头模型(scalp skull brain,SSB)上进行的仿真实验证明了利用两步MREIT算法进行头部组织三维电阻抗图像重构的合理性与可行性。结论该两步MREIT算法可以用于头部组织三维电阻抗图像重构,具有潜在的应用价值。Objective To develop a new Two-step magnetic resonance electrical impedance tomography （MREIT） algorithm based on radial basic function （RBF） neural network for imaging electrical impedance distribution of a head. Methods Firstly, the magnetic resonance imaging （MRI） system with high resolution was used to set up 3D model of the object and to identify the boundaries of different tissues. Then RBF MREIT algorithm was applied to estimate piece-wise homogeneous impedance values of those tissues, respectively. Furthermore, the impedance of each element within each region of the FEM model was estimated according to the RBF genetic algorithm method based on the piece-wise constant impedance. Results Computer simulations were conducted in a three-sphere head model （scalp-skull-brain, SSB） and the simulation results showed the applicability and feasibility of the present Two-step MREIT algorithm in imaging continuous electrical impedance distribution within the head. Conclusion The present Two-step MREIT algorithm is an effective method for imaging the continuous electrical impedance distribution within the human head.

关 键 词：核磁共振电阻抗成像 磁感应强度测量 径向基函数 神经网络 遗传算法 

分 类 号：R319[医药卫生—基础医学]