P299 iCSD may produce spurious results in dense electrode arrays
Joseph Tharayil*1,2,Esra Neufeld2, Michael Reimann1,3 1 Blue Brain Project, École polytechnique fédérale de Lausanne (EPFL) Campus Biotech, Geneva, Switzerland 2Foundation for Research on Information Technologies in Society (IT'IS), Zurich, Switzerland 3Open Brain Institute, Lausanne, Switzerland
*Email: tharayiljoe@gmail.com Introduction
Estimation of the current source density (CSD) is a commonly-used method for processing and interpreting local field potential (LFP) signals by estimating the location of the neural sinks and sources that give rise to the LFP. However, recentin vivoexperiments using dense electrode arrays have found surprising CSD patterns, with high-spatial-frequency oscillations between current sinks and sources [1].
Methods By analytically computing the contribution of a two-dimensional Gaussian current source centered on an electrode array to the CSD as a function of array density, the width of the current source, and the location of the current source (using the standard CSD method [2]), we show that spurious results mistaking true sources for sinks and vice versa are obtained when the inter-electrode spacing is small relative to the current distribution width (Fig. 1a). To study the practical relevance of this issue, we simulated LFP recording in a detailed model of rat cortex (200’000 morphologically detailed neurons, (Fig. 1b) [3]). We estimate CSD from these recordings using the inverse CSD (iCSD) [4] method, and, for a variety of electrode densities and CSD estimation parameters, compare the results to the ground-truth current distribution and to the “non-negative” CSD, a metric similar to standard CSD method but which ignores regions where confounding of sources and sinks occur. Results With high-density arrays, our model of rat cortex produces the same high-spatial-frequency oscillation between sinks and sources observed in [1] (Fig. 1c.i-iv). As array density increases, the correlation between iCSD and ground-truth current density decreases (Fig. 1c). Modifying iCSD parameters improves the correlation, but the correlation between ground-truth CSD and non-negative CSD is consistently better than the correlation between ground-truth CSD and iCSD.
Discussion Our results indicate that the high-spatial-frequency oscillations observed inin vivoCSD calculated using high-density electrode arrays are likely due to confusion between sinks and sources. This confusion occurs because the assumption underlying CSD estimation — that current sources are homogeneous over some radius in the plane perpendicular to the electrode array — is not satisfiedin vivo. While more accurately specifying this radius parameter does improve the CSD estimate, there is no value which results in a better correlation between iCSD and ground-truth than the correlation between non-negative CSD and ground truth, suggesting that the true CSD is not homogeneous at any scale.
Figure 1. Fig. 1: a: A positive current source can produce a negative CSD contribution. b: Model of rat cortex (from [3]). c: Comparison of iCSD and objective CSD for various array spacings. Acknowledgements References [1]http://dx.doi.org/10.7554/eLife.97290 [2]https://doi.org/10.1016/0165-0270(88)90056-8 [3]https://doi.org/10.1101/2023.05.17.541168
