基于多维模式分析对说谎的脑网络特征识别

为了研究说谎时的脑网络特征,采集了32个被试在说真话和说谎条件下的功能磁共振数据,预处理后利用AAL模板构建不同条件下的功能连接网络,再利用基于机器学习的多维模式分类器对说谎和说真话进行分类。该分类器取得了良好的分类正确率82.03%(说谎84.38%,说真话79.69%),并提取了辨别说谎和说真话的有效的功能连接模式。结果表明了使用大尺度的功能连接对说谎和说真话进行分类的良好性能,并且从脑网络角度揭示了说谎的特征。

Brain-Network Feature Recognition of Deception Based on Multivariate Pattern Analysis

Considerable functional MRI (fMRI) studies have shown differences of brain activity between lie-telling and truth-telling. However there are few studies aiming at brain network feature of lie-telling. In this study, we obtained fMRI data of 32 subjects while responding to questions in a truthful, inverse and deceitful manner, then constructed whole-brain functional connectivity networks for the lie-telling and truth-telling conditions based on a canonical template of 116 brain regions, and used a multivariate pattern analysis approach based on machine learning to classify the lie-telling from truth-telling. The results showed that the classifier achieved high classification accuracy (82.03%, 84.38% for lie-telling, 79.69% for truth-telling) and could extract informational functional connectivities that could be used to discriminate lie-telling from truth-telling. These informational functional connectivities were mainly located among networks. These results not only demonstrated good performance when classifying with functional connectivities, but also elucidated the neural mechanism of lie-telling from a functional integration viewpoint.