Source estimation

We review briefly source estimation procedures and Sam Williamson's contributions to them. The original source model of choice was the equivalent current dipole, born both out of computational necessity and philosophical choice. In the last decade, high resolution MRIs have become routinely available, and now automatic segmentation algorithms make possible highly detailed tessellated cortical surfaces for each subject. Thus imaging algorithms mapping directly onto the cortex are now practical, but require strong spatial priors in order to map the hundreds of measurements onto the tens of thousand of cortical locations. Scanning algorithms, also known as beamformers, help to bridge the gap between linear imaging approaches and nonlinear modeling approaches. Thus the modern workflow is to linearly transform complex montages of sensor data onto interpretable cortical images, scan these images for simpler underlying models, and identify plausible source regions and their temporal dynamics.