Multiple incomplete views clustering via non-negative matrix factorization with its application in Alzheimer's disease analysis
Multiple incomplete views clustering via non-negative matrix factorization with its application in Alzheimer's disease analysis
复制标题
DOI：
10.1109/isbi.2018.8363834
复制
发表时间：
2018-04
期刊：
影响因子：
--
通讯作者：
Kai Liu;Hua Wang;S. Risacher;A. Saykin;Li Shen
中科院分区：
文献类型：
--
作者：
Kai Liu;Hua Wang;S. Risacher;A. Saykin;Li Shen
传统的神经成像分析，例如对阿尔茨海默病（AD）收集的数据进行聚类，通常依赖于来自单一成像模式的数据。然而，最近的技术和设备进步为我们提供了更好地分析疾病的机会，我们可以收集和利用来自不同图像和遗传模式的数据，这些数据可能会提高预测性能。为了在AD分析中进行更好的聚类，在本文中，我们进行了一项新的研究，利用来自不同模态/视图的数据。为了实现这一目标，我们提出了一种简单而有效的方法，基于非负矩阵分解（NMF），不仅可以实现更好的预测性能，但也处理一些数据丢失在某些视图。在ADNI数据集上的实验结果证明了该方法的有效性。
Traditional neuroimaging analysis, such as clustering the data collected for the Alzheimer's disease (AD), usually relies on the data from one single imaging modality. However, recent technology and equipment advancements provide with us opportunities to better analyze diseases, where we could collect and employ the data from different image and genetic modalities that may potentially enhance the predictive performance. To perform better clustering in AD analysis, in this paper we conduct a new study to make use of the data from different modalities/views. To achieve this goal, we propose a simple yet efficient method based on Non-negative Matrix Factorization (NMF) which can not only achieve better prediction performance but also deal with some data missing in some views. Experimental results on the ADNI dataset demonstrate the effectiveness of our proposed method.