This PhD work deals with forward and inverse problem in magnetoencephalography (MEG) and electroencephalography (EEG). Three topics are addressed. The forward problem is issued by the boundary element method (BEM). A new formulation, called symmetric formulation, is proposed. This new formulation is then applied to the electrical impedance tomography (EIT) for which two conductivity estimation algorithms are proposed. The inverse problem is addressed in the imaging approach framework. Image regularization techniques by means of diffusion processes are transposed to the inverse problem in order to constrain the reconstruction of distributed sources. Several algorithms are proposed, one of them computes the minimal total variation solution.Ce...