基于自适应标记分水岭算法的肝脏CT图像自动分割

目的为减少人工交互提出了基于自适应标记分水岭的CT系列图像肝脏区域自动分割算法。方法首先对图像进行形态学重构运算以平滑图像,然后计算多尺度形态学梯度,同时提出利用梯度图像非零的局部极小值点的均值进行自适应标记提取,以避免分水岭的过分割和欠分割,再结合肝脏为最大的实质性脏器和相邻图像的相似性实现CT系列图像的肝区自动分割。结果该算法能自动、快速地提取CT系列图像中的肝脏区域。结论分水岭算法能准确定位区域的边缘,通过选择合适的阈值对梯度图像进行标记以抑制分水岭的过分割,实现医学图像中感兴趣区域的自动分割。

Automatic segmentation of liver CT images based on adaptive marked-watershed algorithm

Objective An automatic segmentation method of liver area from computed tomography images based on adaptive marked-watershed algorithm is proposed to reduce human interaction.Methods The image is smoothed by morphological reconstruction,and the morphological gradient is calculated by the multi-scale structuring elements.Then the mean of non-zero local minimum of the gradient images is computed for adaptive Marked,which can overcome the over-segmentation and insufficient-segmentation of the watershed algorithm.Finally,by using the information that the liver is the largest organ and the similarity of the adjacent images,the regions of the liver are automatically extracted from the CT images.Results The region of the liver can be automatically and quickly extracted from the series of CT images by the proposed algorithm.Conclusions The watershed algorithm can accurately locate the edge of the area.By choosing the appropriate threshold as the adaptive Marked for the gradient image to suppress the watershed over-segmentation,the adaptive marked-watershed algorithm can automatically extract the regions of interest from medical images.