Purpose
The middle temporal area (MT) plays an important role in motion processing and receives its most important input signals from the primary visual cortex (V1). Previous receptive field models were not designed to represent the highly specialized class of neurons which project from V1 to area MT; and they have not been able to do so. In particular, previous models do not explain why with ascension up the V1 to MT specialization hierarchy, the proportion of temporal bandpass neurons to temporal lowpass neurons increases (Hawken et al, 1996). We sought to do so here.
Methods
Recently, we introduced the Gabor-Einstein wavelet basis (S.G. Odaibo, Soc. Neurosci 2013 Abstract 259.08). It is a model for representing the receptive fields of neurons projecting from V1 to area MT. Like the Gabor function, the Gabor-Einstein wavelet is the product of a gaussian envelope and a wave carrier. However, unlike the Gabor function, the Gabor-Einstein wavelet's wave carrier is a relativistically-invariant sinc function whose argument is the energy-momentum relation. Here, we run numerical simulations in MATLAB to investigate the response properties of V1 to MT neurons.
Results
We find that in the Gabor-Einstein basis, bandpass temporal filter elements are necessarily composed of lowpass temporal filter elements. Consequently, a greater number of Gabor-Einstein basis elements are generally required to represent neurons on the more specialized (MT) end of the spectrum. This feature is consistent with MT neurons inheriting most of their salient properties from V1 neurons. Plotted below are a Gabor-Einstein wavelet centered at (0,0) and (1.5,1.5) respectively.
Conclusions
Previous receptive field models are not able to represent the highly specialized class of neurons which project from V1 to area MT. In particular, these models do not explain why the proportion of temporal bandpass neurons increases with ascension up the V1 to MT specialization hierarchy. This phenomenon is not arbitrary. Instead, it is a strong clue about the particular spatiotemporal structure of the V1 to MT magnocellular stream, and is naturally represented in the Gabor-Einstein wavelet basis. This model and the experiments it motivates will provide new fundamental insights into how substrates of the motion percept are encoded.
Keywords: 755 visual cortex •
673 receptive fields •
473 computational modeling