Behavioral and functional neuroimaging investigation of motion integration
in early visual cortical hierarchy
Jury:
Edmund Derrington (President),
Anna Montagnini and Rainer Goebel (reviewers),
Mathilde Bonnefond, Thérèse Collins, Rosanne Rademaker, Henry Kennedy and Kenneth Knoblauch (supervisors).
Abstract
Visual perception depends on multiple factors, both external (the information available on the retina) and internal (neural coding and learned expectations based on past experience and context). Notably, the interpretation of partial, noisy or ambiguous visual signals requires inference mechanisms aiming to identify what we perceive as a function of what is the most probable explanation. Thus, it seems that the brain is capable of weighing and comparing multiple scenarios in order to resolve ambiguity inherent in visual input.
In the example of the perception of a moving object, several levels of visual processing have been described. Cells in early visual cortical areas (V1 and V2) exhibit motion sensitivity that is very limited in space, responding to local and relatively simple (1-D) spatio-temporal variations. Higher in the visual cortical hierarchy, more specialized areas (e.g., Medial Temporal, MT/V5, and Superior, MST) are capable of integrating multidimensional motion signals, such as those generated by moving objects in the world, hence allowing coding of more complex motion phenomena. The first two-stage models of motion integration within the visual cortical hierarchy proposed a feedforward processing of information, from V1/V2 (i.e., local 1-D motion detection) to MT/V5 (multi-dimensional integration). Nonetheless, multiple electrophysiological studies showed evidence of the presence of neurons as early as area V1 responding to complex motions in macaques. Moreover, Transcranial Magnetic Stimulation studies in humans demonstrated that MT/V5-to-V1 feedback connectivity was essential for the perceptual awareness of motion. Altogether, these findings suggest a role for MT/V5-to-V1 feedback in motion perception and lead to the prediction that signals related to complex motion integration can be detected as early as area V1.
To test this hypothesis, I performed psychophysical and functional Magnetic Resonance Imaging (fMRI) experiments in humans, using an ambiguous, bistable moving stimulus, composed of two differently oriented drifting gratings, each moving in a different direction. During prolonged viewing, the appearance of this stimulus spontaneously switches between a pair of drifting component gratings transparently sliding across each other and a coherent plaid pattern drifting in an intermediate direction. The alternation of percept between the two semi-stable states (component versus pattern motion) suggests competitive dynamics between two equiprobable interpretations. Our hypothesis was that, with extended viewing of this stimulus, the activity in early visual areas (V1/V2) would reflect the variations of perceived motion, hence in accord with the hypothesis of an active role of feedback connections in resolving ambiguous visual signals. We showed evidence that within all regions of interest (V1, V2, MT/V5), direction-selective subunits reflect the perceptual decisions experienced by the observers, and that this effect is not correlated to any bias in the direction of reflex eye movements. Future research using high-resolution fMRI will enable more precise localization of motion integration signals linked to the perceptual switches, hence allowing better identification of the neural substrates implicated in the cortical interactions described.
