基于序列特性的胸部标记物自动分割

手术导航系统广泛使用外置标记物进行点配准,在配准过程中,需要选取图像中标记物的中心.为了实现标记物中心自动定位,基于肺穿刺手术导航系统设计了一种自动分割标记物算法.该算法先对胸部CT图像初分割得到疑似标记物;然后根据标记物在胸部图像中的序列特性,对疑似标记物进行筛选,得到正确标记物.实验结果表明,该方法与手动分割标记物的重合率达到95%以上,中心坐标误差小于0.22 mm.

Automatic Segmentation of Fiducial Marker Based on Sequence Characteristics

The fiducial markers are widely used for point-pair registration in the image-guided surgery. When performing the registration， the centers of the fiducial markers will be picked. For localization of the fiducial markers， an automatic approach for the segmentation of markers was proposed based on the lung biopsy image-guided surgery system. Firstly， the candidate markers were coarsely segmented from the chest CT image. Then the real markers were detected and confirmed from the candidate markers according to the sequence characteristics approach. Experiments showed that， comparing with the results of manual segmentation， the method can segment the markers with 95% pixel overlap rate and the center localization error is less than 0.22mm.