CNS*2020 Online has ended
Welcome to the Sched instance for CNS*2020 Online! Please read the instruction document on detailed information on CNS*2020.
Back To Schedule
Sunday, July 19 • 7:00pm - 8:00pm
P81: A Computational Neural Model of Pattern Motion Selectivity of MT Neurons

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
https://unimelb.zoom.us/j/99736655097?pwd=VFNuOGozMkpaU2JyeGpKdlpUb0JvQT09    Password: 813336

Parvin Zarei Eskikand
, David Grayden, Tatiana Kameneva, Anthony Burkitt, Michael Ibbotson

The middle temporal area (MT) within the extrastriate primate visual cortex contains a high proportion of direction-selective neurons. When the visual system is stimulated with plaid patterns, a range of cell-specific MT responses are observed. MT neurons that are selective to the direction of the pattern motion are called “pattern cells”, while those that respond optimally to the motion of the individual component gratings of the plaid pattern are called “component cells”. The current theory on the generation of pattern selectivity of MT neurons is based on a hierarchical relationship between component and pattern MT neurons, where the responses of pattern MT neurons result from the summation of the responses of component MT neurons [1]. Where the gratings cross in plaids, the crossing junctions of the gratings move in the pattern direction. However, revealing the ends of the moving gratings (terminators) in human perceptual experiments breaks the illusion of the direction of pattern motion: the true directions of motion of the gratings are perceived.

Here, we propose a biologically plausible model of MT neurons that uses as inputs the known properties of three types of cells in the primary visual cortex (V1): complex V1 neurons, end-stopped V1 neurons (which only respond to the end-points of the stimulus), and V1 neurons with suppressive extra- classical receptive fields. The receptive fields of the neurons are modelled as spatiotemporal filters. There are two types of MT neurons: integration MT neurons with facilitatory surrounds and segmentation MT neurons with antagonistic surrounds [2]. A neuron’s pattern or component selectivity is controlled by the relative proportions of the inputs from the three types of V1 neurons. The model provides a simple mechanism by which component and pattern selective cells can be described; the model does not require a hierarchical relationship between component and pattern MT cells.

The results show that the responses of the model MT neurons are highly dependent on two parameters: the excitatory input that the model neurons receive from the complex V1 neurons with extra-classical RFs and the inhibitory effect of the end-stopped neurons. The results also show experimentally observed contrast dependency of the pattern motion preference of MT neurons: the level of the pattern selectivity of MT neurons drops significantly when the contrast of the bars is reduced.

The presented model solves several problems associated with MT motion detection, such as overcoming the aperture problem and extracting the correct motion directions from crossing bars. Apart from the mechanism of the computation of the pattern motion by MT neurons, the model inherently explains several important properties of pattern MT neurons, including their temporal dynamics, the contrast dependency of pattern selectivity, and the spatial and temporal limits of pattern motion detection.

[1] Kumbhani RD, El-Shamayleh Y, Movshon JA (2015) Journal of Neurophysiology 113, 1977-1988.

[2] Zarei Eskikand P, Kameneva T, Burkitt AN, Grayden DB, Ibbotson MR (2019). Frontiers in Neural Circuits 13, 43.


Parvin Zarei Eskikand

University of Melbourne

Sunday July 19, 2020 7:00pm - 8:00pm CEST
Slot 18