基于马尔科夫随机场的乳腺DCE-MRI图像序列配准

Images Registration of Breast DCE-MRI by Markov Random Field

乳腺DCE-MRI扫描过程中,病人运动等会使图像序列产生运动伪影。消除运动伪影的影响,需要对DCE-MRI时间序列图像进行运动补偿。本文充分利用图像信息和强化率在时空上的先验信息,提出采用基于B样条的自由形变模型描述组织的形变场和自由形变模型控制点的位移场,使用离散马尔科夫随机场进行建模。以基于高斯核的残差复杂度为图像间相似性度量,离散马尔科夫随机场的能量函数采用Fast-PD算法快速优化求解组织形变场。最后将求解的组织形变场对DCE-MRI时间序列图像进行运动补偿。对仿真和真实乳腺DCE-MRI图像序列进行实验,实验结果表明,本方法可达到较高的配准精度。

Dynamic contrast-enhanced magnetic resonance imaging（ DCE-MRI） suffers artifact by patient motion during the imaging procedure. It is necessary to correct patient motion effect by deformation registration for DCE-MRI. To make use of the prior knowledge of enhancement in space-time,we present the deformation field which is described by free deformation model based on B spline and the displacement field of free-form deformation control points,and model by the Markov random field（ MRF） model. We adopt the residual complexity（ RC） based on the Gaussian kernel as the image similarity measure. The optimization of discrete MRF energy is quickly completed by Fast-PD algorithm,which can achieve the solution of deformation field. Finally,the solution of deformation field is used to correct the motion of time sequence images. The experimental results on synthetic and real images demonstrated the proposed method could provide high registration accuracy.