联合蚁群算法和PCNN的脑部MRI图像分割方法  

Ant colony optimization combined with PCNN for brain MRI image segmentation

作　　者：吴骏[1] 孙明明[1] 肖志涛[1] 张芳[1] 耿磊[1] 

机构地区：[1]天津工业大学电子与信息工程学院,天津300387

出　　处：《光电子．激光》2014年第3期614-619,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61102150);天津市科技支撑计划重点项目(12ZCZDGX02100);天津市高等学校科技发展基金(20120805)资助项目

摘　　要：采用蚁群算法(ACO)联合脉冲耦合神经网络(PCNN)的脑部磁共振成像(MRI)图像分割方法。其中利用ACO解决了PCNN参数设置困难的问题,同时能够克服图像的低对比度和噪声对图像分割的影响,实现图像的精确分割。首先利用ACO的全局搜索能力,以图像信息熵与灰度期望值的和作为ACO的目标函数,对PCNN的3个关键参数β、αE和VE进行设定;然后基于PCNN简化模型,结合最大熵值准则对脑部MRI图像进行分割;最后对分割结果进行面积滤波,得到最终的分割结果。实验结果表明,本文方法能够实现脑部MRI图像的自动分割,具有较高的精度和较强的鲁棒性。对于没有噪声的图像,本文方法分割结果的平均正确提取率达到97.0%以上,平均错误提取率达到0.4%以下,平均杰卡德相似系数达到94.8%以上;对于添加了不同级别噪声的图像,本文方法的分割效果也优于FCM和自适应PCNN。A novel method that combines ant colony optimization （ACO） and pulse coupled neural net- work （PCNN） for brain MRI image segmentation is proposed. The ACO is used to solve the problem of parameters setting for PCNN. The influence of low contrast and noise on image segmentation is over- come and accurate segmentation can be achieved. Firstly, the sum of image information entropy and gray mean is used as the target function of ACO and the global search ability of ACO is used to set three key parameters of PCNN. Then the brain MRI image is segmented by the simplified PCNN model which combines the maximum entropy criterion. Finally, the ultimate segmentation result is obtained via area filtering. Experimental results show that the proposed method can segment brain MRI image automati- cally and achieve relatively high accuracy and robustness. For images without noise, the average correct extraction rate of the method is above 97.0% ,the average error extraction rate of it is below 0.4% and the average Jaceard similarity coefficient of it is above 94. 8%. For images added with different levels of noise, segmentation results of the method are better than those of FCM and adaptive PCNN.

关 键 词：医学图像分割 脉冲耦合神经网络(PCNN) 蚁群算法(ACO) 

分 类 号：TN911.73[电子电信—通信与信息系统]