基于散斑纹理评估的超声分割方法

Ultrasound Image Segmentation based on Speckle Texture Assessment

针对传统超声图像分割算法预处理去除散斑噪声时对超声信号造成不可逆转的系统误差问题,提出了一种基于模糊逻辑的不依赖预处理的超声图像分割算法。该算法用一种散斑噪声纹理评估的计算方法来量化超声图像各部分的散斑纹理影响程度,将这种散斑纹理影响值作为分割的一个参考量输入模糊逻辑模型,最终经过模糊逻辑计算实现医学超声图像的分割。基于临床超声图片的实验,该方法所划分散斑区域和经验划分的区域重合率达到92%以上,相对传统的实验方法提高了13%。可见,所提算法在不依赖去噪处理的情况下,提高了分割准确性,并且在散斑纹理形变大的情况下维持了很好的分割效果。

Aiming at the problem that the irreversible system error would be produced on the ultrasound signal with traditional Ultrasound image segmerrtation algorithm for preproeessing and de-nosing, an ultrasound image segmentation algorithm based on speckle assessment and without relying on de-noising is proposed. This segmentation algorithm could quantize the impacts of speckle by measuring speckle texture features, and this value would be taken as a parameter of fuzzy logic model. Finally, segmentation algorithm is implemented with this fuzzy logic model. The experimental results based on clinical images show that this segmentation algorithm can improve the accuracy of segmentation without relying on de-nosing, and even when large deformation is found in speckles.