摘要
医学电阻抗成像(MEIT)技术是一个复杂的非线性问题,图像重建算法对所构图像质量具有重要的影响。由于传统的BP神经网络算法存在训练速度慢且易陷于局部极小等特点,采用全局寻优的遗传算法(GA)实现BP网络连接权的学习,并将该混合算法用于电阻抗断层图像(EIT)重建。实验结果表明,该算法所构图像无论被测场域中心还是边缘处均具有较高的图像分辨率,与典型的反投影及灵敏度系数算法所构图像相比,其质量得到明显的改善。
The medical electrical impedance tomography technique is a complex nonlinear problem. The image reconstruction algorithm affects the image quality seriously. Because there exists drawbacks of slow training speed and easy falling into the local minimal value using the BP algorithm, the genetic algorithm as a global searching optimum which is used to train the connecting weights of the BP neural network is described, and is applied to reconstruct images of conductivity distribution. Experimental results show that images reconstructed by the hybrid algorithm have higher image resolution not only in the center but also at the edge of the measured field. Compared with the LBP algorithm andthe sensitivity coefficient method, the image quality has been improved markedly.
出处
《计量学报》
CSCD
北大核心
2003年第4期337-340,共4页
Acta Metrologica Sinica
基金
国家自然科学基金(39870827)
天津市自然科学基金(013614411)
国家新技术研究发展专项经费(20001AA413210)
关键词
计量学
电阻抗成像
神经网络
遗传算法
Metrology
Electrical impedance tomography
Neural network
Genetic algorithm
