Attentional modulation of receptive fields in a computer model of the thalamocortical system

TitleAttentional modulation of receptive fields in a computer model of the thalamocortical system
Publication TypeConference Paper
Year of Publication2011
AuthorsNeymotin, S. A., Kerr C., Francis J. T., & Lytton WW.
Conference NameSociety for Neuroscience 2011 (SFN '11)
KeywordsSFN, Society for Neuroscience
Abstract

The concept of a nonclassical receptive field (RF) arose with the recognition that brain state profoundly influences the shape of an RF. These influences will be mediated by diffuse neuromodulators projecting from brainstem areas, or by specific projections from other cortical or thalamic areas associated with attention or perceptual integration. We used a computer model of a thalamocortical system to evaluate Layer 4 (L4) RF changes caused by activation from higher integrative and attentional areas projecting to L2/3. Our model consisted of 81 neocortical columns arranged in a 9x9 grid. Each column consisted of 99 event-based integrate-and-fire cells arranged to reflect sensory cortex layers, for a total of 8019 cells. Cells received Poisson input and input from other cells via AMPA/NMDA/GABAA synapses. A 9x9 set of thalamacortical cells (TC) projected to the columns, providing feedforward activation of layer 4 stellate cells (E4). TCs projected to overlap adjacent columns, but with reduced strength. We formed population RFs for each column as the number of L4 excitatory cell (E4 cell) spikes during presentation of a topographic input. We simulated attentional processes by increasing activation to E2 cells. At baseline each column responded maximally to the set of 3x3 TCs projecting directly to it, with fall-off with distance from center of RF. We modeled feedback attentional modulation as activation of L2/3. This produced activation of L4 stellate cells which broadened their classical (spatial) RF and also provided increased duration of response. Threshold for response to topographic inputs was reduced across both granular and infragranular areas (L5). This alteration could not be clearly demonstrated in L2/3 due to the direct activation. Our results demonstrate that the complex neocortical multiplexing being performed through integration of multiple inputs can be partially captured using a broadened conception of RF representation. In addition, we predict that attention can enhance or decrement communication between columns, depending on neuromodulators present in the system. This may allow for a trade-off in processing between higher sensitivity to weak inputs (RF broadening) vs. higher spatial resolution (RF sharpening).