Predicting neuronal dynamics with a delayed gain control model.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual System Modeling

Background:

• Visual neurons exhibit complex response dynamics including sub-additive summation, adaptation to sustained stimuli, and slower responses at low contrast.
• These dynamics are observed across various measurement techniques, including intracranial electroencephalography (EEG), functional magnetic resonance imaging (fMRI), and single-unit recordings.

Purpose of the Study:

• To propose and validate a simple computational model that predicts diverse visual neuronal response dynamics.
• To investigate regularities in visual processing across human visual field maps and species.

Main Methods:

• Developed a computational model incorporating linear filtering, expansive exponentiation, and divisive gain control with a delayed gain signal.
• Input: time-varying stimulus contrast; Output: predicted neuronal dynamics.
• Model fitted to intracranial EEG data from human patients and validated against macaque single-unit spiking and human fMRI BOLD signals.

Main Results:

• The model successfully predicts neuronal adaptation, sub-additive summation, and contrast-dependent response speeds.
• Fitting the model to human intracranial EEG revealed systematic differences in temporal receptive fields across visual areas and faster gain control in later visual areas.
• The model generalized to explain contrast-dependent spiking in macaque V1 and fMRI BOLD amplitudes in human V1.

Conclusions:

• A simple, unified model can capture complex and seemingly disparate visual neuron response dynamics.
• The delayed gain control mechanism is crucial for accurately predicting observed neuronal dynamics.
• The model provides insights into visual processing regularities across human visual field maps and species-specific responses.