Sensory constraints on volitional modulation of the motor cortex

Carmen Fernández Fisac, Steven M. ChasebioRxiv 2023 · DOI

TL;DR

This study reveals that the primary motor cortex (M1), crucial for voluntary movement, is not solely under volitional control. Sensory feedback, particularly the orientation of visual cues, significantly impacts an individual neuron's modifiability, suggesting that sensory context imposes inherent constraints on motor control.

What Problem Does This Paper Address?

The prevailing view in neuroscience, particularly in the context of brain-computer interfaces (BCIs), is that motor cortical neurons can be volitionally modulated with sufficient training and feedback. While studies have shown remarkable flexibility in this modulation, the extent to which non-volitional sensory inputs might constrain this volitional control has remained an open question. Specifically, researchers wanted to understand if the sensory context in which a movement occurs can limit the degree to which M1 activity can be voluntarily controlled, even when the underlying neural task remains the same. This is crucial for understanding the fundamental mechanisms of motor control and for designing more effective BCIs.

Methods

The researchers developed a novel task called the Feedback Alteration Single-cell Task (FAST). In this paradigm, macaque monkeys were trained to control the firing rate of a single neuron in their primary motor cortex (M1), designated as the "command neuron" (CN). This CN's firing rate was directly translated into the real-time position of a computer cursor along a one-dimensional axis. The core of the FAST paradigm lies in its ability to manipulate the sensory feedback associated with this task while keeping the neural requirements identical.

Key methodological aspects include:

  • Single-neuron control: Monkeys learned to modulate the firing rate of individual M1 neurons to move a cursor.
  • Sensory context manipulation: The orientation and location of the cursor's movement axis were systematically varied across different blocks of trials.
  • Controllability metric: A novel metric called "controllability" was developed. This metric quantifies the ease with which a neuron's firing rate can be modulated by assessing the area of a control ellipse, which represents the covariance between the neuron's firing rate and its time derivative. Larger ellipse areas indicate better volitional control.
  • Adaptive difficulty: An automated difficulty schedule was implemented to gradually increase task demands by separating firing rate targets, ensuring that performance was consistently challenged.
  • Focused and multi-day training: To investigate the role of learning, subjects underwent focused training on specific orientations and extended training over multiple days.

Datasets

Not discussed in the paper. The study utilized data collected from three rhesus macaques performing the FAST paradigm.

Key Insights and Ideas

  • Sensory context is a powerful constraint: The most striking insight is that the orientation of the visual feedback axis dramatically impacts the ability to volitionally control individual M1 neurons. What is effortless in one orientation can become problematic in another, even with the same neural task.
  • Learning doesn't fully overcome constraints: Extended training, both within a session and across multiple days, did not eliminate the influence of sensory context on neural controllability. This suggests that these constraints are not simply due to a lack of practice or understanding.
  • Orientation matters more than location: While both axis orientation and location influenced controllability, orientation had a substantially larger effect. This highlights the specific importance of the directional aspect of sensory feedback.
  • Implications for BCIs: The findings suggest that BCI decoder design needs to account for these sensory-neural relationships. Building decoders that are consistent with existing sensory interactions in the population might be crucial for learnability and performance.

Key Results

  • Orientation dependence: A majority of tested neurons (96% in Monkey N, 100% in Monkey R, 80% in Monkey A) showed significant differences in controllability across different movement axis orientations. On average, orientation explained 35.8% of the variance in controllability across animals.
  • Location effects: Location of the movement axis also affected controllability, but to a lesser extent. On average, location explained 14.6% of the variance in controllability across animals.
  • Training insensitivity: Focused training on a "difficult" orientation improved controllability at that orientation but also generalized to other orientations, without eliminating the overall orientation dependence. Multi-day training also failed to abolish the interaction between orientation and controllability.
  • Controllability changes: While training did not eliminate sensory dependence, it did lead to an increase in overall controllability across all orientations.

Limitations and Future Work

  • Single sensory modality: The study focused solely on visual feedback. The authors acknowledge that other sensory modalities (e.g., proprioception, auditory cues) could also influence volitional modulation. Future work could explore these interactions.
  • Hard-wired constraints vs. learning: The persistent sensory dependence raises two possibilities: either there are hard-wired constraints that are difficult to unlearn, or subjects simply didn't discover the optimal mental strategy. Further research is needed to disentangle these possibilities.
  • BCI decoder design: The findings suggest that future BCI development should consider how to leverage or mitigate these sensory-neural interactions for improved performance and learnability.
  • Internal states: The authors mention that internal states like attention and emotion could also affect controllability, which warrants further investigation.

Summary updated

Abstract

ABSTRACT Voluntary movement is driven by the primary motor cortex (M1), and individuals can learn to modulate even single neurons at will. Yet M1 also receives pronounced sensory inputs and contributes to sensory-driven motor responses. To what extent do these non-volitional signals restrict voluntary modulation of M1? Using a task in which the firing rate of a single neuron directly determines the position of a computer cursor along a visual axis, we assessed the ability of monkeys to modulate individual neurons under different sensory contexts. We found that sensory context persistently affected volitional control of single neurons in M1. For instance, visually rotating the biofeedback axis could render the same neural task effortless or problematic. Notably, extended training within or across days did not resolve this disparity. Our findings suggest that sensory context can limit the degree to which M1 activity is under volitional control.
BibTeX
@misc{fisac2023sensory,
  author = {Carmen Fernández Fisac and Steven M. Chase},
  title = {{Sensory constraints on volitional modulation of the motor cortex}},
  year = {2023},
  doi = {10.1101/2023.01.22.525098},
  howpublished = {bioRxiv},
}

Loading related content...

Figure