随着肺癌发病率的增高和影像技术的进步,越来越多的微小肺癌(尤其是肺部磨玻璃结节)得以检出,对于此类病变首先考虑手术治疗.但肺是含气组织,在术中会发生萎陷.巨大的体积变化导致肺内结节也随之发生位移,术前CT确定的位置在术中无法定位.针对这一问题,在术中肺萎陷动物模型基础上,提出了一种融合3D-SIFT特征的B样条配准方法.该方法首先提取局部特征,筛选并匹配较好的特征点;然后依据浮动图像与参考图像匹配的特征点,得到浮动图像的初始位置;最后利用控制网格对浮动图像进行配准.经初步研究,对于不同萎陷状态的肺部图像,应用此方法得到了很好的配准结果.因此,融合3D-SIFT特征的B样条配准方法适合肺部图像配准,将来有望应用于微小肺癌的术中定位.%With the increased incidence of lung cancer and the progress of medical imaging technology,more and more small pulmonary nodules ( especially for ground glass nodules) have been detected. Surgical treatment is the first consideration for such nodules. But lung is a gas containing organ,which will collapse during operation. With the huge volume change,pulmonary nodule will change its position, which makes location difficult in the operation. To solve it,an animal model database simulating lung collapse in operation is established. A B-spline curves registration method with fusion of 3D-SIFT features is proposed. Local feature is extracted firstly,and then the feature points are selected and matched. According to the feature points the initial position of the floating image is obtained. Used the control grid,the floating image is registered finally. For a series of lung CT images in different collapse states,satisfactory registration results are obtained by it which is suitable for pulmonary images registration and can be used for intraoperative localization of small pulmonary nod-ules in the future.
展开▼
