基于Snake模型的低对比度噪声心脏MRI图像分割

提出了一种基于Snake模型的低对比度、噪声心脏MRI图像分割算法。Snake模型通过轮廓线的变形得到感兴趣区域(ROI)的边界，但其分割结果依赖于初始轮廓线的位置，且变形曲线容易收敛于局部梯度极大值区域或从弱边界处泄漏，对于深度凹陷的区域也难以分割。通过增加局部面积能量项，扩大了Snake模型寻找边界的范围；在模糊C均值集群分类的基础上，构造模糊能量项，能够较好地处理心脏MRI图像中的弱边界、局部梯度极大值区域、伪影等现象。分割实验证明了改进的模型能够有效地分割低对比度、噪声心脏MRI图像。

Low Contrast Noisy Cardiac MRI Image Segmentation Based on Snake Model

A segmentation algorithm of low contrast noisy cardiac MRI based on Snake model is proposed. The snake model can find the boundary of Region of Interest (ROI) by deforming the spline curve, but the segmentation result relies on the initial location of the curve and it is easy for the curve to converge to the local gradient maximum region or leak from the weak edges. Moreover, the model cannot segment the concave region accurately. The improved Snake model can search for the edge of ROI in the wider region by adding local area energy term. A fuzzy energy term is added to the model for dealing with weak edges, local gradient maximum region and artifacts in the MRI. The segmentation experiments demonstrate the effectiveness of the algorithm listed in the paper for the low contrast noisy cardiac MRI.