作　　者： ; ; ;

机构地区： 湖南大学电气与信息工程学院

出　　处： 《电子测量与仪器学报》 2008年第4期26-29,共4页

摘　　要： 磁共振图像的准确分割对于辅助医生确定病灶的位置和形状、制订治疗方案和评价治疗效果具有重要的意义。本文提出了一种新的磁共振图像(MRI)分割方法。构造了一种自组织小波神经网络,通过融合T1、T2和Pd图像的特征来识别MRI中生物组织的类别。该网络继承了小波分析局部精度高和神经网络自学习能力强的优点,采用自组织算法利用训练数据的稀疏性对网络的结构和初始参数进行优化,简化了网络结构,提高了网络学习的速度,避免了网络陷入局部最优学习。将所提方法应用于大脑磁共振图像分割的实验结果表明,所设计的自组织小波神经网络MRI图像分割方法具有精度高和学习速度快的优点。 Correct segmentation of magnetic resonance image （MRI） is very important for doctors to ascertain the shape and position of the focus, prepare cure scheme and evaluate cure effect. A new segmentation method of MRI is proposed in this paper. A self-organization wavelet neural network is designed for recognizing different tissues in MRI by fusing the features of T1, T2 and Pd images. The proposed wavelet neural network inherits high local accuracy of wavelet analysis and strong self-learning ability of neural network. A self-organization algorithm is used to optimize the structure and initial parameters of the network. This can simplify the network structure, improve the network learning speed, and avoid the local optimization of network learning. The experiments applying the proposed method to the segmentation of brain MRI were carried out. Experiment results show that the proposed segmentation method has high precision and fast learning speed.

关 键 词： 小波 神经网络 磁共振图像 图像分割 自组织

领　　域： [生物学] [自动化与计算机技术] [自动化与计算机技术]