Yuhui Du*, Xingyu He, and Vince D Calhoun. A New Semi-Supervised Non-Negative Matrix Factorization Method For Brain Dynamic Functional Connectivity Analysis. In 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI), pp. 1591-1594, 2021.
时间：2021-08-07 10:13:20 来源： 点击：
To overcome the shortcoming of static brain functional connectivity analysis, recent studies have analyzed brain functional connectivity from a time-varying view to identify subtle group differences as potential biomarkers. However, how to obtain reliable functional connectivity states from dynamic functional connectivity (DFC) is challenging, since the resulting states are often sensitive to the cluster or component number. In our present work, we propose a new semi-supervised non-negative matrix factorization method combining with a community detection technique, which can automatically estimate reliable functional connectivity states from DFC without a need of setting a cluster or component number. We applied our method to fMRI data of 36 schizophrenia patients and 49 healthy controls to investigate brain connectivity changes in schizophrenia. We found significant and meaningful group differences in four of five connectivity states. Our findings suggest that our method can help understand the mechanism of mental illness in a more effective way.