基于投影近似不变性的3D-2D医学图像配准

3D-2D Medical Image Registration Based on Near Projective Invariance

研究了同一病人的MRA和DSA图像之间的3D-2D配准问题,提出了一种基于投影近似不变性的配准算法,即在同一视角下,同一病人的三维MRA血管骨架的投影与二维DSA的骨架保持一致。通过定义二者之间的代价函数,并利用牛顿迭代法可得到视角参数的最优解。针对临床采集的MRA和DSA数据进行了实验,取得了较为满意的结果。

The registration between MRA and DSA of a same patient is studied in this paper. The registration algorithm based on near projection invariance is presented, i.e. the projection of 3D MRA skeleton should accord with the 2D DSA skeleton under the same perspective view. After the cost function between the projections is defined, the view angle can then be obtained through maximizing the cost function using Newton iteration method. Satisfactory results are obtained in the clinical MRA and DSA data of the same patient.