UC San Diego

Beamformer adaptive spatial filters have been used extensively in the field of magnetoencephalography (MEG) as tools to reconstruct functional activation of the brain. Conventional single beamformer techniques suffer from distortion in the presence of coherent activation of the cortex or are difficult to use due to the need of a priori information. These qualities present a major disadvantage to analyzing human brain responses, as coordinated functional responses require a degree of synchronous activation in different parts of the active cortex. In this dissertation, a novel beamformer technique, the multi-core beamformer, is developed that is robust to source correlation and does not require the use of a priori information. This novel approach is tested in both simulated and real experiments, including auditory and median-nerve stimulation, which provide well-studied systems to gauge the effectiveness of our new technique. Simulations show that the multi-core beamformer can successfully determine source time-courses, source powers, and source locations while minimizing or eliminating the distortion present in other methods. Results from real- life experiments show that the multi-core beamformer produces physiologically meaningful solutions that agree with previous functional imaging and neurophysiology studies. The use of the multi-core beamformer is expected to greatly contribute to the analysis of MEG recordings and, in general, improve our understanding of functional brain activity