Adaptive temporal integration of motion in direction-selective neurons in macaque visual cortex

Area of Science:

• Neuroscience
• Visual Perception
• Computational Neuroscience

Background:

• Direction-selective neurons in primary visual cortex (V1) and extrastriate motion area MT/V5 are crucial for motion perception.
• These neurons integrate spatiotemporal information, but how this integration depends on stimulus properties is not fully understood.
• Existing models often rely on linear filter mechanisms, which may not capture the complexity of neural responses.

Purpose of the Study:

• To investigate the dependence of temporal integration in direction-selective neurons on stimulus speed, spatial frequency (SF), and contrast.
• To compare neuronal responses with predictions from a standard motion energy model.
• To elucidate the nonlinear mechanisms underlying cortical motion processing.

Main Methods:

• Recorded from direction-selective neurons in V1 and MT/V5 of the visual cortex.
• Used randomly moving sinusoidal gratings with varying speed, SF, and contrast.
• Analyzed neuronal responses using spike-triggered average (STA) analysis and compared them to a linear filter-based motion energy model.

Main Results:

• Temporal integration windows varied significantly with stimulus parameters: longer for slow motion, high SF, and low contrast.
• At low speeds and high SF, larger STA peaks indicated greater information per spike when mean firing rates were low.
• Observed trends were consistent across V1 and MT, correlating with psychophysical data but not explained by the motion energy model.

Conclusions:

• Cortical motion processing in V1 and MT is highly nonlinear and stimulus-dependent.
• Simple linear filter models are insufficient to explain the output of direction-selective neurons.
• Spike rate tuning may overlook crucial aspects of motion coding, especially at low firing rates.