肠癌内镜图像中病变的细微特征捕获模型仿真  

Simulation on Subtle Feature Capture Model of Lesions in Intestinal Cancer Endoscopic Image

作　　者：王哲[1] 李海瑛[1] 张钰[1] 

机构地区：[1]河北医科大学临床学院,河北石家庄050017

出　　处：《计算机仿真》2015年第6期320-323,355,共5页Computer Simulation

摘　　要：肠癌内镜图像中病变区域的细微特征的捕获,对于肠癌早期的发现和诊断具有重要的意义。肠癌内镜图像中病变区域的细微特征对光感图像回应的特征较弱,无法在CT图像中形成较为明显的回波特征。传统的捕获模型捕获方法在对弱特征捕获时,很容易当成噪声特征过滤,造成捕获准确性降低,提出一种采用局部多尺度梯度变换算法的肠癌内镜图像中病变区域的细微特征捕获方法。计算肠癌内镜图像中病变区域的细微特征不同尺度,利用非最大抑制法提取细微病变特征点。将内镜图像转换为只有病变区域的细微特征与背景的二极值图,利用质心亚像素法对病变区域的细微特征点的坐标进行重新定位,消除回波特征较弱的干扰,提高了识别的精确度。实验结果表明,改进算法能够有效提高肠癌内镜图像中病变区域的细微特征捕获的准确性。The subtle features capture of lesions area in intestinal cancer endoscopic image, has important signifi- cance for early detection and diagnosis of intestinal cancer. The characteristics that subtle features of lesions area in the intestinal cancer endoscopic image respond to the light image are weaker, which cannot form the obvious echo fea- tures in CT images. A subtle features capture method of the lesion area in the intestinal cancer endoscopic image was proposed based on local multi - scale gradient transform algorithm. Different scales of subtle features of the lesion ar- ea in the intestinal cancer endoscopic image were calculated, and the feature points of subtle lesions were extracted u- sing non - maximum suppression method. The endoscopic image was converted to bi - extremal graph only with subtle features and background of the lesion area. The coordinate of subtle feature points in lesion area was re - positioned using centroid sub - pixel method, to eliminate the interference of weak echo characteristics and improve recognition accuracy. The experimental results show that, the improved algorithm can effectively enhance the accuracy of the subtle features capture of lesion area in the intestinal endoscopic image.

关 键 词：肠癌 内镜图像 细微特征 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]