Dept of Physiology, Anatomy and Genetics, Univ of Oxford, Oxford, UK
A surprisingly small number of studies have addressed the fundamental question to what extent the anatomical properties of macaque visual cortex can explain the physiological characteristics of simple cells in layer 4C. Based on well-established anatomical data, we created a realistic computational model of the thalamocortical pathway in macaque to determine whether the classical Hubel and Wiesel model for constructing orientation-tuned simple cells from LGN inputs can account for the physiological properties observed in extracellular studies of the primary visual cortex (V1).
Some of the major anatomical constraints, taken from the literature, in our model are the extent of the axonal arborisation of LGN axons in cortical layer 4C (for P-cells, 0.2 mm diameter on average, for M cells 0.6 mm on average and 1.2 mm maximum), the spread of horizontal connections in layer 4C (projections up to 3-4 mm in upper 4C alpha), the spread of the dendritic field of spiny stellate neurons in layer 4C (maximum 0.2-0.25 mm diameter) the variation in cortical magnification factor as a function of eccentricity (E) in V1, the variation in the density of LGN neurons with E, and the number of LGN P and M cells that project to a single layer 4C neuron (on the order of 10-30). The major physiological constraints in the model are the classical receptive field (CRF) sizes of V1 and LGN neurons measured with patches of drifting sinusoidal gratings, the extra-classical surround suppression in the LGN, the orientation tuning bandwidth and the spatial frequency preference and bandwidth in V1.
The model was built and tested within a flexible framework that allowed it to operate on arbitrary time-vary binocular achromatic visual stimuli. It consisted of networks of tens to hundreds of thousands of neurons, depending on E and size of visual field being tested, and included Poisson or conductance-driven integrate-and-fire spiking mechanisms.
We found that for E between 2-8 deg, RFs of average or greater size are anatomically implausible under the classical Hubel and Wiesel model. In particular, assuming RFs are built from M LGN afferents, direct feedforward connections can account only for CRF sizes up to about 1.5 deg at E=8 deg and up to about 0.5 deg at E=2 deg, whereas physiological RF sizes ranging from 0.5 to 3 deg are common. For P-cell afferents, the validity of the classical model is almost completely abolished. By including lateral projections within layer 4C, we found that a more realistic range of CRF sizes could be explained.