Related Experiment Videos
Distributed population mechanism for the 3-D oculomotor reference frame transformation.
Michael A Smith1, J Douglas Crawford
1York Centre for Vision Research, Canadian Institute of Health Research Group for Action and Perception, Department of Psychology, York University, Computer Science Building, 4700 Keele Street, Toronto, Ontario M3J 1P3, Canada.
Journal of Neurophysiology
|November 13, 2004
Summary
This study trained a neural network to generate accurate eye movement commands, revealing a distributed population mechanism for transforming visual information into motor signals, likely mirroring how the brain performs this visuomotor transformation.
Area of Science:
- Computational Neuroscience
- Oculomotor Systems
- Machine Learning in Biology
Background:
- Human saccades necessitate a complex reference frame transformation from visual input to motor commands.
- Primate neurophysiology suggests this transformation occurs between the superior colliculus and brain stem burst neurons, but the mechanism remains unclear.
Purpose of the Study:
- To elucidate the neural mechanisms underlying the eye orientation-dependent reference frame transformation for saccades.
- To model how the brain might achieve accurate 3-D saccade generation using a neural network.
Main Methods:
- Trained a 3-layer neural network using a 2-D visual map and eye position signals as input.
- Network outputted eye orientation displacement commands to drive simulated brain stem burst neurons.
- Analyzed the internal representations and computational properties of the trained network's hidden units.
Main Results:
- The neural network successfully generated kinematically correct 3-D saccades, including orientation-dependent motor command adjustments.
- Hidden units developed eye-centered visual receptive fields and "gain field"-like eye position sensitivities.
- A distributed population mechanism was identified, where subtle adjustments in unit contributions rotated the population motor vector.
Conclusions:
- The trained network demonstrated a plausible neural mechanism for visuomotor reference frame transformation.
- This distributed population coding approach likely represents the solution employed by the biological oculomotor system.
- The findings offer insights into how the brain integrates visual and eye position information for motor control.