Abstract: In order to improve the accuracy of segmentation，reduce the complexity of computation and enhance the level of automatic segmentation，a new method for automatic segmentation of proximal femur is presented.The trusted contour of proximal femur can be obtained by bidirectional segmentation，making use of the characteristics of the circle-like femoral head and middle section and high similarity in adjacent sections of femur in CT images.The results of the experiments show that the proposed method can accurately extract the contour of proximal femur from CT images of hip joint，which has practical value.

Key words: femur, CT images, image segmentation

摘要： 针对现有髋关节序列CT图像中骨骼近端分割精度差、自动化程度低、计算复杂度高的问题，提出一种自动分割股骨近端的方法。该方法充分利用序列CT图像中股骨头与股骨中段呈现类圆形以及相邻切片中股骨特征相似度高的特点，融合双向分割的股骨轮廓信息得到可靠的股骨近端轮廓。实验结果表明，该方法能准确地从髋关节序列CT图像中提取出股骨近端的轮廓信息，具有一定的实用价值。

关键词: 股骨, CT图像, 图像分割