Conclusions We have presented an algorithm that reduces the complexity of localizing multiple neural sources by exploiting the time dependence of the data. We have shown that on a realistic head model with simulated EEG data, our algorithm is capable of correctly predicting the number of independent sources in the model and reconstructing potentials due to each source separately. These potential maps can be successfully used by source localization methods to independently localize sources. By integrating our algorithm within the SCIRun problem solving environment, we were able to computationally steer the multi-start downhill simplex algorithm towards probable regions of activation. Interactive control of the simulation, coupled with statistical data preprocessing of the data enabled us to dramatically increase the efficiency and accuracy of recovering multiple sources from EEG data.     Fri Oct 8 13:55:47 MDT 1999

 
Revised: March , 2005