Object recognition by selective spike and LFP data in inferior temporal cortex

Gabriel Kreiman, Chou Hung, Alec Shkolnik, Tommy Poggio, Jim DiCarlo

Society for Neuroscience, San Diego, 2004

How accurately can we read out information about objects from neuronal activity in the brain? We recorded single electrode data, including multi-unit activity (MUA) and local field potentials (LFP) from the inferior temporal cortex of macaques while they passively viewed 78 different pictures of complex stimuli. As reported previously, we observed that spike counts form MUA showed selectivity for some of the pictures. Strikingly, the LFP data, which is thought to constitute an average over large numbers of neurons, also showed statistically significant selective responses. Typically, the LFP responded selectively to less than 10% of the presented pictures and showed less selectivity than MUA data. Control experiments from shuffled data and also from recordings outside the inferior temporal cortex did not show selectivity. Our preliminary observations suggest that there was no strong overlap between the selectivity of MUA and LFP recordings from the same electrode. To attempt to decode the information from spikes and LFP data, we applied a classifier to disriminate between 8 possible groups of stimuli. Assuming independence between different sites, we estimate how classification performance changes with the number of sites, and other parameters including the bin size and integration time. Using both the MUA and LFP data as input to the classifier improves the performance compared to either data source alone. This observation suggests that the information carried by the LFPs is not redundant with that carried by MUA. Selectivity from MUA activity at a given site was correlated with the selectivity observed at a site separated by approximately 200-800 mm.. This observation, together with the LFP data suggests that there is some topographical arrangement to the organization of selectivity in inferior temporal cortex.