脑电是睡眠分期的重要依据，但功能磁共振的超高磁场环境严重降低了脑电信号的质量，成为开展睡眠神经影像研究的瓶颈。随着越来越多的实验室装配磁共振系统，开展脑电结合磁共振的研究，亟待发展相应的脑电信号处理和分析方法，来拓宽电生理技术在神经影像上的应用范围。本项目拟借助贝叶斯机器学习的最新进展，发展一系列适用于高强磁场环境的睡眠脑电分析技术。主要包括：1)开发适合睡眠监测的环境噪声记录线圈，借助肌电、眼电、胸式和腹式呼吸张力等信号，提高脑电伪迹的去除效率；2)将磁场环境噪声的统计特征纳入到脑电源定位模型中，结合动态功能磁共振分析，开发适合于超高磁场环境的脑电源定位技术；3)利用贝叶斯线性判别分析，引入脑电皮层网络特征，开发基于磁共振平台的自动睡眠分期算法。本项目的最终目标是开发出适合于功能磁共振的多导睡眠图系统，满足睡眠认知神经科学研究实验的需要。

EEG is the most important information for sleep staging. However, the strong fMRI noise severely reduced the quality of EEG signal, and this has become a bottleneck in the sleep imaging. As increasing laboratories are equipped with the ultra-high MRI system, it is urgent to develop the corresponding EEG signal processing technology to broaden the application of electrophysiology in the field of neuroimaging. This project intends to develop a series of EEG analysis technologies to deal the high noise in the ultra-high magnetic field, with the employment of the latest developed Bayesian Machine Learning. This project mainly includes: 1) Designing noise reference accessories based on carbon wire loop to monitor sleep. With the signal from the EMG, EOG, thoracic and abdominal breathing-tension signals, we will improve the efficiency of EEG artifacts correction. 2) Extending the statistical characteristics of magnetic field noise to the EEG source localization model. After integration with the dynamic analysis of fMRI, we will develop some new source location methods which are suitable for ultra-high magnetic field. 3) Developing an automatic sleep stage classifier based on Bayesian linear discriminant analysis and the features of the cortical network. Our ultimate goal is to develop a polysomnography system that is suitable for the ultra-high magnetic field, which can meet the needs of sleep cognitive neuroscience research.