ETWR: An EEG/MEG Source Reconstruction Method Using Spatio-Temporal Two-Way Regularization

• Activity Number: 76
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Tian Siva Tian*+ and Jianhua Huang and Haipeng Shen and Zhimin Li
• Companies: University of Houston and Texas A&M University and The University of North Carolina at Chapel Hill and Medical College of Wisconsin
• Keywords: Two-way regularization ; Spatio-temporal data ; EEG/MEG ; Inverse problem ; Source reconstruction

Abstract

This paper presents a spatio-temporal two-way regularization method for EEG/MEG source reconstruction. The proposed method, referred to as ETWR, belongs to the category of ``imaging methods' which estimate the dipole locations and amplitudes simultaneously by penalized least squares. ETWR produces reconstructed sources with several desirable properties, namely, temporal smoothness, spatial focality and smoothness. The temporal smoothness is guaranteed by imposing a roughness penalty on the temporal domain of the estimated source time courses. In the spatial domain, a sparsity-inducing penalty is imposed to ensure the focality, and a discretized Laplacian operator is employed to ensure that the amplitudes transit smoothly among the neighboring dipoles at each time point. An efficient, multilevel block coordinate descent algorithm is developed for computation. Simulations and an MEG example show the effectiveness of ETWR.