一种基于先验信息的脑功能网络提取方法*

目的 脑功能网络的提取在脑科学研究中具有重要意义，本文提出一种基于先验信息的脑功能网络提取方法。方法该方法首先基于先验信息得到初始的目标和背景种子点，然后基于图论将整个脑图像构建图，最后利用半监督聚类技术提取脑功能网络。基于不同信噪比的模拟数据，本文对提出方法、基于种子点的方法、独立成分分析方法、以及两种聚类方法（归一化最小化割和K均值方法）进行比较。基于真实脑静息态功能核磁共振数据，本文使用提出方法对默认模式网络进行提取。结果 基于模拟数据的实验结果表明提出算法相对于传统的方法可以得到更为准确且鲁棒的脑功能网络。基于静息态功能核磁共振数据得到的默认模式网络在一些重要脑区具有高的稳定性，且不同地点采集数据得到的结果具有较强的一致性。结论 提出方法是一种有效的脑功能网络提取方法。

A method for brain functional network extraction based on prior information

Objective Extraction of brain functional networks is of great importance for brain research. This paper proposed a prior information based method to identify brain functional networks. Method The method firstly obtained objective and background seeds based on prior information, and then constructed a graph for the brain image using graph theory, finally extracted the brain functional network using a semi-supervised clustering technique. Based on the simulated data with different signal to noise ratios, the proposed method, the seed based method, independent component analysis, and two clustering methods (Normalized cuts and K-means) were compared. Based on the real resting-state functional magnetic resonance imaging (fMRI) data, default mode network was extracted using the proposed method. Result Simulation based experimental results show that compared to the traditional methods, the proposed method can obtain more accurate and robust brain functional network. Real fMRI based experimental results illustrate that the default mode network had high stability in important regions as well as great consistency among data from different sites. Conclusion The proposed method is an effective method for brain functional network extraction.