基于动力学聚类与α散度测度的动态心肌PET图像因子分析

发布时间：2018-04-13 12:22

本文选题：因子分析 + α散度　；参考：《南方医科大学学报》2017年12期

【摘要】：目的建立一种基于动力学聚类与α散度测度的因子分析方法,分析从动态心肌PET图像中无创地提取血输入函数及局部组织的时间活度曲线。方法通过最小化动态图像与因子模型间的α散度将动态PET图像做初步的分解,得到初始因子与因子图像,然后联合PET像素动力学聚类的先验信息解决因子分析模型中解的非唯一性问题,将初始因子与因子图像通过空间变换生成具有生理意义的组织活度曲线及组织空间分布。结果与传统的最小二乘法测度和最小化因子图像间重叠程度约束模型相比,本模型对噪声的敏感性较低,提取出的结果的精确性较高。结论通过选取最优的α值作为因子分析模型的测度,并引入PET图像像素的动力学聚类信息,能精确地获得血输入函数及局部组织的时间活度曲线,在视觉评价及量化评价均具有优质表现。
[Abstract]:Objective to establish a factor analysis method based on dynamic clustering and 伪 divergence measure to analyze the time activity curve of blood input function and local tissue from dynamic myocardial PET images.Methods by minimizing the 伪 divergence between the dynamic image and the factor model, the dynamic PET image was preliminarily decomposed to obtain the initial factor and factor image.Then combining the priori information of PET pixel dynamic clustering to solve the non-uniqueness problem of the solution in the factor analysis model, the initial factor and the factor image are transformed into the physiological tissue activity curve and the tissue spatial distribution by spatial transformation.Results compared with the traditional least square measure and the minimum factor image overlap degree constraint model, this model is less sensitive to noise, and the accuracy of the extracted results is higher.Conclusion by selecting the optimal 伪 value as the measure of the factor analysis model and introducing the dynamic clustering information of the pixels of the PET image, the blood input function and the time activity curve of the local tissue can be accurately obtained.In visual evaluation and quantitative evaluation have high quality performance.
【作者单位】：南方医科大学生物医学工程学院//广东省医学图像处理重点实验室;
【基金】：国家自然科学基金(81501541,61628105) 广东省自然科学基金(2014A030310243,2016A030313577) 国家重点研发计划(2016YFC0104003) 广州市珠江科技新星专项(201610010011)~~
【分类号】：R541;R817.4

【相似文献】