基于局部特征与轮廓控制的磁共振扩散张量插值方法

MR Diffusion Tensor Image Interpolation Method Based on Local Features and Profile Controlling

目的在保持生物体扩散张量物理性质和控制张量行列式轮廓基础上进一步提高张量插值的质量。方法提出一种基于局部特征与轮廓控制扩散张量图像轮廓控制插值方法,也即基于邻近多个张量值的对数-欧氏距离,对基于黎曼测地线上非匀速运动的轮廓控制插值方法的空间距离进行修正。结果脑部磁共振扩散张量图像的插值重建实验结果表明,本文方法和其他常用插值方法相比,3种在不同的误差度量下误差较其他方法均有显著性改善(P<0.01),任取样本的各项异性分数值误差也有一定的减小。结论在脑磁共振扩散张量图像插值中,本文方法能保持生物体扩散张量物理性质并控制行列式轮廓,减少扩散张量插值的误差,改善插值效果。

Objective To improve the quality of diffusion tensor imaging （DTI） interpolation based on control- ling the interpolation determinant profile and keep the diffusion tensors＇ physical properties at the same time. Methods An improved interpolation scheme based on local features and profile controlling was proposed. That is to say, considering the Log-Euclidian distance of some nearby tensors, our scheme fixed the space distance of the interpolation based on non-unifoml motion on a Riemannian geodesic. Results The interpolation recon- struction experiment of medical DTI images showed that the interpolation error of our scheme had significant improvement （P 〈 0. 01 ） using different error metrics. And the fractional anisotropy error of tensors in random- ly selected samples was reduced. Conclusion For the interpolation of brain DTI image, our scheme can keep the diffusion tensors＇ physical properties and control the interpolation determinant profile. The error of DT1 in- terpolation can be reduced so that the interpolation is improved.