基于主动形变模型的医学序列图像的分割

对主动形变模型(active shape model)方法进行了改进,将其用于CT序列图像的分割,实现了对一些边缘不清晰的软组织的分割。一个组织在CT序列图像中其形状变化往往是连续的,据此,我们先在序列图像中按顺序每间隔几幅抽取一幅图像,用其他方法分割好;然后对分割结果使用主成分分析方法(Principle Component Analysis)计算其形变特征,生成可变形轮廓模板;再用这个模板去变形、配准、分割其他图像,从而达到一个比较理想的分割结果。

Algorithm for Segmentation of Medical Image Series Based on Active Shape Model

An algorithm based on the active shape model was proposed for the soft tissue segmentation of medical image series. The soft tissue edges in medical image are blurry, so the soft tissue segmentation is very difficult. But its shape usually deforms sequentially in medical image series. An image was extracted every several images from the image sequence and segmented them by the other methods. The segmented images were used as the training maps of an active shape model. The model was created by Principle Component Analysis Algorithm, so this model could only deform in ways characteristic of the tissue it represented. This model was used to segment the left images in the image sequence by mutual information registration and active contour model. The experiments show the efficiency of this method.