时空模型结合模拟退火进行脑磁源的定位

Spatio-Temporal MEG Source Localization Using Simulated Annealing

脑磁源的定位问题是脑磁图 (m agnetoencephalography,MEG)研究的一个基本问题。其中多偶极子定位是脑磁逆问题研究当中的难点。本文通过研究脑磁图的时空模型 STSM(spatio- temporal source modeling) ,提出将时空模型与模拟退火相结合进行多偶极子的定位 ,以克服其他优化方法易落入局部极小的不足。时空模型中偶极子参数经分解可分为线性部分和非线形部分 ,只对非线性部分进行模拟退火优化大大降低了优化空间的维数。通过与 MUSIC(MU ltiple SIgnal Classification)方法的比较 ,发现将时空模型与模拟退火相结合可以相对降低对源信号独立性的要求。

Locating the sources of brain magnetic fields is a basic problem of magnetoencephalography (MEG). The locating of multiple current dipole is a difficult problem for the inverse study of MEG. A method combining Spatio-Temporal Source Modeling with Simulated Annealing to locate multiple current dipoles, is presented through studying the STSM of MEG. This method can overcome the shortcoming of other optimal methods to avoid being trapped in a local minimum. The dipole parameters can be separated into linear and nonlinear components. The optimization dimensions can be reduced greatly by just optimizing the nonlinear components only. Compared with the MUSIC (MUltiple Signal Classification), this method can cut down requirements of independence of the dipole sources correspondingly.