Electro/Magnetoencephalography Signal Processing Methods and Performance

脑电/脑磁图信号处理方法和性能伊利诺伊大学芝加哥分校脑电图学系检测大脑中的电源对于了解其功能和临床应用都很重要。例如，在手术治疗前绘制大脑活动图和寻找癫痫活动灶。我们正在开发检测方法，通过对头部周围传感器阵列的测量数据进行计算机处理来找到来源。更具体地说，我们使用了电/磁脑图(E/MEG)传感器来测量头皮上的电位和头部外部的感应磁场。我们正在开发几种处理E/MEG信号的新方法，分析它们的性能并用实际数据验证它们的适用性，从而有助于改进E/MEG设备的使用和性能，并增加神经数据处理工具的能力。我们希望解决一些目前最相关的E/MEG问题：(I)估计和跟踪功能和神经元连接的路径，跟踪脑源的轨迹；(Ii)在存在未知时空协方差的噪声存在的情况下，估计集中源和扩展源；(Iii)同时估计源参数和组织电导，(4)开发计算高效的真实感头部模型方法，减少对分割算法的要求；(5)估计具有非均匀历元的诱发反应的源参数。我们还在为评价新提出的方法制定业绩衡量标准，以便与现有系统和技术进行比较；确定哪些方法是有效的，并帮助优化未来系统的设计。最后，我们使用了经验数据集评估和验证方法。这些数据集是由杰弗里·莱文博士的团队从临床和认知神经科学研究中得出的，在这些研究中，全头脑磁图和高密度脑电被同时记录下来。Nehorai小组正在开发处理方法，两个小组将在评估和验证方面进行合作。

Electro/MagnetoencephalographySignal Processing Methods and PerformanceArye NehoraiEECS DepartmentUniversity of Illinois at ChicagoDetecting electric sources in the brain is important for both understanding its function and for clinical applications. Examples include mapping the brain activities and finding foci of epilepsy activities before surgical treatment. We are developing detection methods that find the sources through computer processing of measurements from arrays of sensors around the head. More specifically, we employ Electro/Magnetoencephalography (E/MEG) sensors that measure electric potentials on the scalp and induced magnetic field outside the head. We are developing several new methods of processing the E/MEG signals, analyzing their performance and validating with real data their applicability, thus contributing to improvements in the use and performance of E/MEG equipment and to increase the capabilities of neurological data processing tools.We hope to solve some of the most currently relevant E/MEG problems: (i) estimating and tracking paths of functional and neuronal connectivity, following the trajectories of cerebral sources, (ii) estimating concentrated and extended sources, in the presence of noise with unknown spatio-temporal covariance, (iii) simultaneously estimating source parameters and tissue conductivities, (iv) developing computationally efficient methods for realistically-shaped head models, which reduce the demands on segmentation algorithms, (v) estimating source parameters for evoked responses with inhomogeneous epochs. We are also deriving performance measures for evaluating the newly proposed methods allowing comparison with existing systems and techniques; identifying those that are effective and helping in the optimum design of future systems. Finally, we are using empirical data sets evaluate and validate methods. These data sets are being derived by the Dr. Jeffrey Lewine's group from Clinical and Cognitive Neurosciences studies where whole-head MEG and high-density EEG are recorded simultaneously. The Nehorai group is developing the processing methods and the two groups will collaborate on their evaluation and validation.