一种基于边缘检测的医学图像增强方法  

Medical Image Enhancement Based on Edge Detection

作　　者：赖思渝[1] 王宇峰[1] 王娟[2] 

机构地区：[1]川北医学院医学影像学系,四川南充637000 [2]西华师范大学计算机学院,四川南充637002

出　　处：《电脑与信息技术》2013年第2期13-16,共4页Computer and Information Technology

摘　　要：现在是用数字图像进行临床诊断的时代。文章提出了一种结合非采样轮廓波变换(NSCT)和脉冲耦合神经网络(PCNN)优点的肿瘤检测方法。该方法首先分别对一组正常人的脑部CT和MRI图像及一位40岁酗酒男性的脑部MRI和PET图像施行三次样条插值配准,并进行非采样轮廓波变换获取其高频和低频信息。将低频子带系数输入PCNN神经元经计算获得融合图像低频系数,对于高频部分对比度被用于激化PCNN网络。最后经逆NSCT变换生成融合图像,并将该图像用Canny算子进行边缘检测。结果显示第一组的融合图像中高密度组织得到了增强并减少了像素扭曲且肿瘤组织能被检测,第二组的融合图像清晰显示了脑部解剖结构同时壳核、尾状核也到得了明确定位。由于非采样轮廓波变换优良的方向性和几何表达能力,该方法能够为外科医生提供精确的肿瘤定位方案。In this paper, a combined method of non-subsampled contourlet transform （NSCT） and pulse coupled neural network （PCNN） is proposed to detect tumor. Firstly, a cubic spline interpolation approach is performed to register source images, these images are decomposed by using NSCT in order to obtain high frequency and low frequency coefficients. Then low frequency information is input to PCNN network and the low frequency coefficients of fusion image are output, as for high frequency part the contrast is used to motivate PCNN. At last, the fusion image is generated by conducting inverse NSCT and is used to detect edges in Canny operator mode. Results show that the spatial property of fusion image is improved and as the excellent directivity and geometric representation ability of the NSCT, the method provides surgeons precise tumor localization scheme.

关 键 词：断层扫描 核磁共振 非采样轮廓波变换 脉冲耦合神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]