利用自适应纹理分布的活动形状分割前列腺磁共振图像

Segmentation of prostate magnetic resonance image with active shape of adaptive texture distribution

基于前列腺磁共振图像性质,提出利用自适应纹理分布的活动形状图像分割方法来自动分割前列腺磁共振图像。该方法首先通过图像的分类与拟合确定感兴趣的腺体区域,同时估计若干形状参数用于分割过程中调整形状;然后融入多重纹理信息,建立纹理一致测度,将传统的活动形状按照自适应的纹理判别步骤细分为纹理分布形状与补充形状,提高活动形状的搜索匹配能力。在搜索匹配部分,利用已估计参数优化活动形状搜索的初始估计,并根据纹理分布形状和补充形状调整迭代过程。实验结果表明,该方法分割出来的前列腺轮廓与金标准的Hausdorff距离为6.00 pixel,分割精度为93%。该方法对活动形状的改进是有效的,利用自适应纹理分布的活动形状能够自动、准确地将前列腺从磁共振图像中分割出来。

On the basis of properties of magnetic resonance images for the prostate, an active shape im- age segmentation method making use of adaptive texture distribution was introduced to segment a prostate magnetic resonance image. Firstly, a prostate region of interest was determined through im- age classification and image fitting, and several shape parameters were estimated and used in the seg- mentation. Then, multi-features were fused to build a texture coincidence measure. In order to im- prove the searching and matching ability of an active shape, the active shape was divided into two por- tions, the texture distribution shape and the supplementary shape. In search, the estimated parame- ters were used to optimize the initial estimation of the actiw~ shape searching and adjust the iterative process based on the texture distribution shape and the supplementary shape. Experimental results in- dicate that the Hausdorff Distance is 6.00 pixels between the true prostate contour and that extracted by the proposed method and the segmentation accuracy of the method is 93%. The proposed methodcan modily the active shape effectively, and can automatically segment the prostate magnetic reso- nance images with high enough accuracy.