我們提出了一個在最小範數法的架構內，從腦磁/電圖估算腦可能的活動區域的方法。這個方法將整個大腦以格點覆蓋，每個格點和少量格點組合求最小範數解，利用少量格點的組合之中源的解會大於多數非源格點解的性質，經由統計每個格點在多次不同的組合中解大於每個組合半數格點之解的次數，篩選出可能是源的格點，再對篩選出的格點進行源分布迭代(SIMN)定位源的位置。數值模擬顯示本方法大幅提升源定位正確率，用人臉辨識腦磁圖數據對本方法進行測試，結果和fMRI的結果一致。;A method in the framework of minimum norm estimate(MNE) is proposed to identify the possible active regions of the brain from MEG/EEG data. The whole brain is covered with grid points. Each grid point is teamed up with a small number of grid points to calculate the MNE of these points. In a group of few grid points, the MNE of source grid point is in general greater than that of non-source grid points. By considering the ensemble of each grid point teaming up with few grid points of a fixed number, the statistic of the number of times that each grid point’s MNE is greater than half of the grid points’ MNE in each combination enables us to identify the possible source grid points. SIMN is applied to these grid points to determine the positions of sources. Numerical simulations show that the accuracy of source localization is greatly improved. This method is applied to the facial recognition MEG data and the results are consistent with the results of fMRI.
