基于时空连续约束的4D脑图像分割模型

4D Brain Image Segmentation Model Based on Spatio-Temporal Information Continuity

纵向分析脑解剖结构的变化可以预测脑组织的生长或萎缩状态,为临床治疗和科学研究提供必要的依据.但由于成像设备或模式不同以及成像时间间隔较长,3D的分割方法得到的结果无法体现脑组织在时间维上缓慢变化的特征.针对这一问题,提出一种基于时空约束的4D脑图像水平集分割模型.该模型包含了由全局以及局部信息组成的数据拟合项、空间平滑项以及时间平滑项.其中数据拟合项体现了各个时间点的图像灰度信息,空间和时间平滑项则能保证分割结果在时空维上体现其缓慢变化的特性.实验结果表明本文方法既能保证准确的分割结果又能保证空间维以及时间维上的连续性.

Longitudinal analysis of brain anatomical change can predict the growth or atrophy of human brain and provide a necessary foundation for clinical medicine application and research.However,due to different imaging machine or model and a long time interval of each image in different time point,the 3D image segmentation method can not provide adequate longitudinal stability of brain tissue variation.In this paper,we propose a 4D brain image level set segmentation model based on spatio-temporal information continuity.This model contain three terms：data term created by global and local information,spatial and temporal smooth term respectively.The data term reflects the intensity information of the image in each time point.The spatial and temporal term can keep the segmentation results smooth variation in these two dimensions.The experiments demonstrate that the proposed method can obtain a temporally consistent and spatially adaptive longitudinal brain image segmentation results.