复杂网络主成分分析的分类方法在音乐家白质可塑性研究中的应用

发布时间：2018-05-07 15:55

本文选题：主成分分析(PCA) + 支持向量机(SVM)　；参考：《中国生物医学工程学报》2015年02期

【摘要】：人脑在多种时间和空间尺度上都是复杂网络,而复杂网络中往往包含着大量的连接信息。主成分分析(PCA)方法主要被用于从大量信息中提取重要特征,因而可以被用于探寻复杂网络中的一些重要信息。众所周知,音乐家是研究训练导致的脑可塑性问题的一个理想模型,探求音乐家脑网络的可塑性变化是非常有意义的。首先通过基于弥散加权成像(DWI)数据的纤维束追踪,构建了16位音乐家与16位非音乐家的脑白质结构网络;然后对两组人的整体脑网络进行了PCA分析,进而对得到的每个主成分做支持向量机(SVM)分类处理,得到分类效果最好的主成分;最终找出对此主成分贡献前1%的连接即为音乐家相对于非音乐家在大脑白质结构网络上发生改变的主要连接。本方法为组间复杂网络对比分析提供了一种基于PCA分类的新思路。基于上述思路,对于音乐家与非音乐家的脑白质结构网络对比分析,表明音乐家在运动、听觉、情绪和记忆等功能脑区表现出更高的脑区间信息传递效率;进而扩展了在网络层面对长期音乐训练改变音乐家白质可塑性问题的理解。
[Abstract]:The human brain is a complex network in many time and spatial scales, and complex networks often contain a lot of connection information. The principal component analysis (PCA) method is mainly used to extract important features from a large amount of information, so it can be used to explore some important information in complex networks. It is well known that musicians are an ideal model for studying brain plasticity caused by training. It is very meaningful to explore the plasticity of musicians' brain networks. Firstly, the white matter structure network of 16 musicians and 16 non-musicians was constructed by fiber bundle tracing based on diffusion-weighted imaging (DWI) data, and then the whole brain network of two groups was analyzed by PCA. Then, every principal component is classified by support vector machine (SVM), and the principal component with the best classification effect is obtained. Finally, the first 1% connection to this principal component was found to be the main connection between musicians and non-musicians in the white matter structure network of the brain. This method provides a new method based on PCA classification for the comparison and analysis of complex networks between groups. Based on the above ideas, a comparative analysis of the white matter structure network between musicians and non-musicians shows that musicians exhibit higher transmission efficiency in brain regions of motor, auditory, emotional and memory functions. It also extends the understanding of long-term music training to change the white matter plasticity of musicians at the network level.
【作者单位】：电子科技大学生命科学与技术学院神经信息教育部重点实验室;
【基金】：国家自然科学基金(91232725,81201159) 中央高校基本科研业务费项目(ZYGX 2011J097)
【分类号】：R338;R445.2

【参考文献】