Control of cortical population activity with patterned microstimulation.

Area of Science:

• Systems Neuroscience
• Neurotechnology
• Computational Neuroscience

Background:

• Closed-loop control of cortical activity is crucial for systems neuroscience and clinical neuromodulation.
• Current methods often require unattainable in vivo circuit models or inefficient trial-and-error open-loop stimulation.

Purpose of the Study:

• To introduce REACHable manifold Control (REACH-Ctrl), a novel data-driven brain-computer interface for real-time control of population spiking activity.
• To demonstrate precise and sample-efficient cortical control using patterned microstimulation and multi-electrode recordings without explicit circuit knowledge.

Main Methods:

• REACH-Ctrl learns a finite-horizon controllability map from short training epochs of random microstimulation and recorded responses.
• It identifies the reachable manifold of population states to compute low-current stimulation sequences for steering neural activity.
• The approach was tested in macaque prefrontal cortex using multi-electrode recordings and patterned microstimulation.

Main Results:

• REACH-Ctrl achieved high accuracy in controlling population spiking activity, demonstrating robustness across sessions and stimulation parameters.
• Geometric analyses revealed that control sequences traverse a well-defined reachable manifold, with partial overlap with the intrinsic neural activity manifold.
• Encoding models indicated that population responses are well-approximated by linear sums of localized stimulation fields, explaining the linear control approach's success.

Conclusions:

• REACH-Ctrl enables precise, sample-efficient control of cortical population activity with clinically relevant hardware.
• This data-driven approach provides a general blueprint for designing effective perturbations in sparsely observed neural circuits.
• The findings advance the potential for closed-loop neuromodulation and brain-computer interfaces.