论文部分内容阅读
目的为分割有偏移场的脑部磁共振图像,建立了一种多相变分水平集的脑部磁共振图像分割与偏移场矫正的耦合模型。方法依据偏移场特性,定义了基于局部灰度信息的K均值聚类准则,并将聚类准则项整合为变分水平集函数的基本能量项,构造变分水平集的能量函数,求解能量函数的欧拉-拉格朗日方程。将提出的变分水平集算法扩展为多相,最终实现脑组织分割和偏移场的矫正。结果以医生手动分割的脑组织为标准,将本文模型与多相LBF模型的分割结果进行定量分析,结果表明,偏移场强度为20%时,本文模型的分割精度比多相LBF模型提高了10%,且随着偏移场强度进一步增强,本模型的优势更明显。结论本模型抗噪能力强且分割结果受参数影响小,可应用于脑组织的分割和偏移场的矫正。
Aim To segment magnetic resonance images of the brain with offset fields, a coupled model of brain magnetic resonance image segmentation and offset field correction was established. Methods According to the characteristics of the bias field, a K-means clustering criterion based on local gray information is defined. The clustering criterion is integrated into the basic energy items of the variational level set function, and the energy function of the variational level set is constructed to solve the energy Euler-Lagrange equations of function. The proposed variational level set algorithm is extended to multiphase, and ultimately the brain tissue segmentation and offset field correction. Results Based on the manual segmentation of brain tissue, the segmentation results of this model and the multiphase LBF model were quantitatively analyzed. The results show that the segmentation accuracy of this model is higher than that of the multiphase LBF model at 20% 10%, and the strength of the migration field is further enhanced, the advantage of this model is more obvious. Conclusion This model has strong anti-noise ability and the result of segmentation is less affected by the parameters. It can be applied to the segmentation of brain tissue and the correction of the offset field.