Conceptual and computational challenges in massive multielectrode data analysis
Extracellular potential recorded in the brain typically reflects activity of multiple cells and processes happening in multiple spatial and temporal scales, depending on the type of electrode and geometric relation between the setups and the tissue. While easy to record, it is notoriously difficult to interpret due to the long range of the electric field. When multiple recordings are available it is possible to estimate the local distribution of current sources (CSD analysis). Our recent method, kernel Current Source Density, allows to estimate CSD from arbitrary distribution of contacts, however, when the number of contacts is large conecptual and computational problems make the CSD analysis difficult. Discuss the challenges appearing in CSD analysis of multielectrode recordings in complex setups. Reproducible kernel Hilbert spaces, singular value decomposition, Python. A wavelet-style multiscale approach to the CSD analysis leads to optimal use of high density probes. For data coming from single cells, morphological information allows one to obtain estimates of CSD distribution along the cell morphology. It is possible to combine recordings of different type, such as ECoG and SEEG, to improve localization of specific phenomena. Kernel CSD analysis and its variants may significantly improve understanding and interpretation of extracellularly recorded brain activity. FINANCIAL SUPPORT: This work was supported by the Polish Ministry for Science and Higher Education grant 2948/7.PR/2013/2 and Narodowe Centrum Nauki grant 2015/17/B/ST7/04123.