波谱学杂志 ›› 2024, Vol. 41 ›› Issue (3): 304-314.doi: 10.11938/cjmr20243094

• 研究论文 • 上一篇    下一篇

基于新一代脑磁图的语义视听单试次检测

郭旭1,2, 王晨旭1,2, 张欣2,3, 常严2,3, 崔峰2, 郭清乾2,3, 胡涛2,3, 杨晓冬1,2,3,*()   

  1. 1.徐州医科大学 医学影像学院,江苏 徐州 221004
    2.中国科学院苏州生物医学工程技术研究所,江苏 苏州 215163
    3.季华实验室,广东 佛山 528200
  • 收稿日期:2024-01-04 出版日期:2024-09-05 在线发表日期:2024-01-19
  • 通讯作者: *Tel: 18900616030, E-mail: xiaodong.yang@sibet.ac.cn.
  • 基金资助:
    季华实验室项目-新一代可穿戴脑磁图仪研制(X190131TD190)

Semantic Audiovisual Single-trial Detection Based on the New Generation of Magnetoencephalography

GUO Xu1,2, WANG Chenxu1,2, ZHANG Xin2,3, CHANG Yan2,3, CUI Feng2, GUO Qingqian2,3, HU Tao2,3, YANG Xiaodong1,2,3,*()   

  1. 1. School of Medical Imaging, Xuzhou Medical University, Xuzhou 221004, China
    2. Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou 215163, China
    3. Jihua Laboratory, Foshan 528200, China
  • Received:2024-01-04 Published:2024-09-05 Online:2024-01-19
  • Contact: *Tel: 18900616030, E-mail: xiaodong.yang@sibet.ac.cn.

摘要:

为解码人脑在语义情境下的视听双模态与单模态中的响应差异,本研究设计了相关任务范式并应用新一代脑磁图(OPM-MEG)结合机器学习方法对采集信号从行为学响应、事件相关场(ERF)和单试次检测3个角度进行分析.结果显示单模态语义响应主要集中在枕叶,而双模态语义响应主要集中在顶叶.同时,双模态下的被试响应速率及单试次检测准确率显著高于单模态.此外,支持向量机(SVM)在4种机器学习模型中显示出了最佳分类效能,在被试内分类平均准确率可达75.16%,被试间平均准确率达80.56%.结果表明基于OPM-MEG结合机器学习为实现解码语义情境下的视听双模态与单模态响应差异提供了一条新的有效途径.

关键词: 新一代脑磁图, 语义, 视听双模态, 机器学习, 事件相关场

Abstract:

In order to decode the difference between audiovisual bimodal and unimodal responses of the human brain in semantic context, this study designed a related task paradigm and applied a new generation magnetoencephalogram combined with the machine learning model to analyze the collected signals from three perspectives: behavioral response, event-related field (ERF) and single-trial detection. Results show that the unimodal semantic response was mainly concentrated in the occipital cortex, while the bimodal semantic response was mainly concentrated in the parietal cortex. At the same time, respondents' response rate and the detection accuracy of single-trial in bimodal mode were significantly higher than that in unimodal mode. Moreover, the support vector machine (SVM) showed the best classification performance among the four machine learning models, with an average classification accuracy of 75.16% for within-subject classification and 80.56% for between-subject classification. This research concludes that the combination of optically pumped magnetometer-magnetoencephalography (OPM-MEG) and machine learning model provides an efficient approach to decode the difference between audiovisual bimodal and unimodal responses of the human brain in semantic context.

Key words: OPM-MEG, semantic, audiovisual bimodal, machine learning, event-related field

中图分类号: