四维CT肺图像提供了肺部动态信息,可广泛用于肺部疾病诊断和精确放射治疗.提出一种针对四维CT肺图像的多层次时空一致性配准方法.该方法首先从最大吸入相位的图像中提取控制点,利用结构张量跟踪算法对控制点进行跟踪;构建多层时空一致性优化模型,对控制点进行调整,从整图配准角度提高控制点配准精度;每次调整后根据配准的一致性误差,添加新的控制点,进一步降低配准的一致性误差,并对新的控制点集重新进行优化调整.该模型在空间上对形变进行约束,保持形变扭曲的平滑性;在时间上对控制点轨迹进行约束,保持控制点的轨迹平滑性;通过一致性约束,降低了正向和反向形变函数的一致性误差.在测试数据集上的实验结果表明,该方法可以得到平滑、稳定、合理的形变结果,可以用于估计肺部运动模型.%Four-dimensional CT lung images provide lung dynamic information,which can be widely used in the diagnosis of pulmonary diseases and precise radiotherapy.This paper proposes a hierarchic spatio-temporal consistent registration for 4D CT lung images.The proposed method extracts control points from the image at the maximum inhalation phase,and tracks these control points by using the structure tensor tracking algorithm;a hierarchical spatiotemporal consistent optimization model is proposed to adjust the control points finely,which has the advantages of improvement of the registration accuracy as a whole;new control points are selected from the region where consistent errors are large,and are added to the control point set after each optimization.These newly added control points are optimized further to reduce the consistent errors of registration.In the proposed method,the spatial smoothness constraint keeps the deformation smooth;the temporal smoothness constraint keeps the trajectories of points stable;the consistent constraint reduces the consistent errors introduced by the transformations in two directions.The experimental results of public dataset show that the proposed method can estimate smooth,stable,and reasonable deformation results.
展开▼
