基于脑连接网络的阿尔茨海默病临床变量值预测

发布时间：2018-03-28 17:19

本文选题：大脑功能　切入点：特征选择　出处：《智能系统学报》2017年03期

【摘要】：目前脑功能连接网络已被广泛用于大脑疾病诊断,然而传统的脑网络分类方法无法评估疾病所处的阶段以及预测病情的发展。近期的研究表明,脑疾病的临床变量值可以有效地帮助医生进行疾病评估,为此提出一种基于脑连接网络的方法,用于对阿尔茨海默病临床变量值进行预测。首先从脑影像中提取功能连接网络,然后使用LASSO进行特征选择,剔除不具有判别性的边。同时融合网络的聚类系数和边的权重作为特征。最后使用支持向量回归机预估临床变量值。在ADNI数据集上对提出的方法进行验证,实验结果表明,提出的方法不仅能够准确地预测疾病临床变量值而且还验证了多种特征融合的有效性。
[Abstract]:At present, brain functional connectivity network has been widely used in the diagnosis of brain diseases. However, traditional brain network classification methods can not assess the stage of disease and predict the development of disease. Recent studies have shown that, The clinical variable value of brain disease can effectively help doctors to evaluate the disease. Therefore, a method based on brain connection network is proposed to predict the clinical variable value of Alzheimer's disease. Firstly, the functional connection network is extracted from brain image. Then use LASSO for feature selection, At the same time, the clustering coefficient and edge weight of the network are fused as the features. Finally, the support vector regression machine is used to predict the clinical variables. The proposed method is verified on the ADNI dataset. The experimental results show that, The proposed method can not only accurately predict the clinical variables of disease, but also verify the validity of multiple feature fusion.
【作者单位】：南京航空航天大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(61422204,61473149) 江苏省杰出青年基金项目(BK20130034) 高等学校博士学科点专项科研基金课题(20123218110009) 南京航空航天大学基本科研业务费项目(NE2013105)
【分类号】：O212.1;R749.16

【相似文献】