Item

Abstract

Local Field Potentials (LFPs) are a complex signal which captures multiple neural contributions, from dendrosomatic dipoles generated by synaptic activity to non-synaptic slow activity such as voltage-dependent membrane oscillations and spike afterpotentials. LFPs exhibit a broadband spectral structure which is traditionally partitioned into distinct frequency bands, initially introduced in the human EEG literature, which are thought to originate from different types of neural events triggered by different processing pathways, such as sensory pathways or neuromodulation. However, the exact frequency boundaries of these processes are not known and, as a result, the frequency bands are often selected based on intuition, previous literature and visual inspection of the data. To address the problem of the arbitrariness of band selection, we defined and implemented numerically a rigorous method to define the number of LFP frequency bands and their boundaries. The criterion chosen for setting the boundaries is to maximize the information about an external correlate carried by the LFP partition. The number of bands is set as the minimum number of bands necessary to provide all the information carried by the LFPs. We applied the partitioning method to the LFPs recorded from primary visual cortex of anaesthetized macaques, and we determined the optimal band partitioning of the [1-250 Hz] LFP range that describes the encoding of naturalistic visual stimuli. We started by partitioning the LFP range into two bands and we successively increased the number of bands in the partition. Four bands (three optimal boundaries) seemed to be enough to extract most stimulus information carried by the LFPs. The first optimal boundary was in the range 50-60 Hz. It partitioned the LFP response into two components - the low frequency range and the gamma range - which have been found (Belitski et. al. 2008) to convey independent information about the natural movie correlate. The second optimal boundary was between 90 and 110 Hz. It subdivided the gamma range into low- and high- gamma frequencies, consistent with recent reports (Gieselmann and Thiele 2008, Ray and Maunsell 2010) that low and high gamma signals may reflect distinct neural processes. The third and last boundary was at approximately 25 Hz. It subdivided the <50 Hz LFP range into two components, one (<12 Hz) highly informative about the visual stimulus and one which was found not to correlate with the visual stimulus.