The data recorded from a cortical column with a linear array electrode contains activity from several neural populations, but the contributions from individual populations are difficult to assess. Barth and collaborators have pursued such an assessment for the low frequency part of the recorded potential in a series of papers. The separation of the potential into population activity was based on principal component analysis (PCA). The PCA expresses the electrode data through spatial loadings (populations) and temporal scores (the temporal activity of the populations). The contribution to the potential from one population can therefore be expressed as the product of its score and loading, and the full extracellular potential is thus a sum of contributions from all populations.
However, the PCA is a purely mathematical method which defines both its scores and loadings as orthogonal vectors, with the first component spanning the maximum variance of the whole data set, the second component spanning the maximum variance remaining and so forth. Instead of using these purely mathematical constraints of the model, we introduce a method where physiological constraints are applied. As a consequence, the expansion will not only be compatible with the physiology, but each component will also have a clear physiological interpretation. This new method is called laminar population analysis (LPA).
Gaute T. Einevoll, Klas H. Pettersen, Anna Devor, Istvan Ulbert, Eric Halgren and Anders M. Dale. Laminar Population Analysis: Estimating ring rates and evoked synaptic activity from multielectrode recordings in rat barrel cortex. J Neurophysiol, 97(3):21742190, Mar 2007.