Electroencephalography (EEG) and magnetoencephalography (MEG) measurements can be used to monitor neural activity, that is generally characterized using current or magnetic dipole source models with time-varying amplitude, position, and moment parameters. The EEG/MEG measurements, however, often contain artifacts that do not originate from the brain. These artifacts can include patient movement, normal heart electrical activity, muscle and eye movement, or equipment and environmental clutter. In this paper, we propose a novel neural activity estimation approach that integrates particle filtering with the probabilistic data association filter in order to validate neural measurements and suppress artifacts before estimating neural activity. Simulations using synthetic data with this approach demonstrate high performance in suppressing artifacts and tracking neural activity; results for real data are also presented.