遗传神经网络在颅脑磁共振图像分割中的应用  

Application of Genetic Algorithm Neural Networks in the Segmentation of Brain Magnetic Resonance Imaging (MRI)

作　　者：黄永锋[1] 赵俊[1] 庄天戈[1] 章鲁[2] 

机构地区：[1]上海交通大学生物医学工程系,上海200030 [2]上海第二医科大学计算机系,上海200025

出　　处：《上海交通大学学报》2004年第5期771-774,共4页Journal of Shanghai Jiaotong University

摘　　要：采用全局最优化算法——实数遗传算法改进BP神经网络的学习算法.在此基础上,提出了一种新的基于实数遗传神经网络的磁共振图像分割算法.实验结果表明,新算法可以有效地降低噪声和边缘模糊对分割结果的影响.另外,设计了一个新的加速遗传算子,可以提高实数遗传算法的收敛速度.The real genetic algorithm, a global maximum method, was applied to improve the learning algorithm of BP. Furthermore, a new algorithm of magnetic resonance imaging (MRI) segmentation based on real genetic neural networks (GNN) was proposed. The experimental results show that the new algorithm is more robust against edge blur and noise. In addition, a new accelerating genetic operator was introduced to improve the convergence of GNN.

关 键 词：颅脑磁共振图像 分割算法 遗传神经网络 实数遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]