论文部分内容阅读
针对肝癌消融计算机断层扫描(CT)图像分割中肿块区域存在成分多变和弱边界问题,为准确提取肝肿块轮廓,提出了一种改进Chan-Vese模型的水平集算法。利用肝与肿块的高斯均值、标准差有显著差异的特点,通过高斯混合模型区分目标与背景的像素隶属,结合边缘梯度信息驱动的长度与形状约束项构造能量泛函,以肿块先验知识确定目标的初始轮廓,促使活动轮廓收敛在目标区域边缘。通过肝CT图像实验数据集验证算法,实现肝上已灭活或部分灭活的癌组织和碘油沉积等构成复杂轮廓提取,实验结果表明,算法平均相似度值大于0.87,其周密性与精确度均优于局部Chan-Vese和局部二值拟合模型。
To solve the problem of variable composition and weak boundary in tumor mass in CT image segmentation for liver cancer ablation, a level set algorithm is proposed to improve Chan-Vese model in order to accurately extract the contour of liver mass. The Gaussian mixture model is used to distinguish the pixel membership of the target and the background, and the energy and functional terms are constructed based on the length and shape constraint terms driven by the edge gradient information. The priori knowledge of the mass is used to determine the Gaussian mean and standard deviation of the liver and mass. The initial outline of the target causes the outline of the activity to converge at the edge of the target area. Through the experimental data set of liver CT image validation algorithm to achieve the liver has been inactivated or partially inactivated cancer tissue and lipiodol deposition constitute a complex contour extraction, experimental results show that the average similarity of the algorithm is greater than 0.87, its precision and precision Degrees are better than the local Chan-Vese and local binary fitting model.