Particle Swarm Optimization Applied to EEG Source Localization of Somatosensory Evoked Potentials

Y. Shirvany, Q. Mahmood, F. Edelvik, S. Jakobsson, A. Hedström, M. Persson. IEEE Transactions on Neural Systems and Rehabilitation Engineering, January 2014, Vol. 22 (1), 11-20.


One of the most important steps in presurgical diagnosis of medically intractable epilepsy is to find the precise location of the epileptogenic foci. Electroencephalography (EEG) is a noninvasive tool commonly used at epilepsy surgery centers for presurgical diagnosis. In this paper, a modified particle swarm optimization (MPSO) method is used to solve the EEG source localization problem. The method is applied to noninvasive EEG recording of somatosensory evoked potentials (SEPs) for a healthy subject. A 1 mm hexahedra finite element volume conductor model of the subject’s head was generated using T1-weighted magnetic resonance imaging data. Special consideration was made to accurately model the skull and cerebrospinal fluid. An exhaustive search pattern and the MPSO method were then applied to the peak of the averaged SEP data and both identified the same region of the somatosensory cortex as the location of the SEP source. A clinical expert independently identified the expected source location, further corroborating the source analysis methods. The MPSO converged to the global minima with significantly lower computational complexity compared to the exhaustive search method that required almost 3700 times more evaluations.

Index Terms—Electroencephalogram (EEG) source localization, finite element method (FEM), inverse problem, magnetic resonance imaging (MRI), particle swarm optimization, somatosensory evoked potential (SEP), subtraction method.


The authors would like to thank Prof. M. Elam from the Sahlgrenska University Hospital for his help to measure the data and S. Bergstrand from the Sahlgrenska University Hospital for all his kind help during the result analysis as well as being subject for the measurement. The authors would like to thank to Assoc. Prof. J. Gellermann for her manual segmentation of the MR images.

Authors and Affiliations

  • Y. Shirvany, Department of Signals and Systems, Chalmers University of Technology and MedTechWest Center
  • Q. Mahmood, Department of Signals and Systems, Chalmers University of Technology and MedTechWest Center
  • F. Edelvik, Fraunhofer-Chalmers Research Centre
  • S. Jakobsson, Fraunhofer-Chalmers Research Centre
  • A. Hedström, Sahlgrenska Academy
  • M. Persson, Department of Signals and Systems, Chalmers University of Technology and MedTechWest Center

Photo credits: Nic McPhee