Dynamic range adaptation in primary motor cortical populations

R Rasmussen, Andrew B. Schwartz, Steven M. ChaseeLife 2017 · DOI

TL;DR

Neurons in the motor cortex, like those in sensory areas, can dynamically adjust their sensitivity to better encode information. This study shows that when monkeys control a brain-computer interface (BCI) in a 3D environment versus a 2D one, their motor cortical neurons change their "tuning" – how they respond to different movement directions. This suggests dynamic range adaptation is a fundamental neural mechanism, not just limited to sensory processing.

What Problem Does This Paper Address?

Sensory neurons, such as those in the visual or auditory systems, are known to exhibit "dynamic range adaptation." This means they adjust their sensitivity to incoming stimuli to maintain optimal information transmission across varying input levels. For example, your eyes adjust to both dim and bright light. The question this paper addresses is whether this same adaptive mechanism exists in the motor cortex, which controls movement, and if it applies to encoding movement direction rather than sensory intensity. Specifically, the authors investigate if motor cortical neurons change their response properties when the dimensionality of the task changes, from 3D to 2D.

Methods

The researchers trained two rhesus macaques to control a cursor in a virtual reality environment using a brain-computer interface (BCI). This BCI setup allowed them to directly link neural activity to cursor movement, bypassing the complexities of natural arm movements (like gravity and inertia).

  1. BCI Control: Monkeys used their motor cortical activity to move a cursor in either a 3D or 2D virtual space. This was achieved using population vector algorithms (PVA) or optimal linear estimators (OLE) to decode neural signals into cursor velocity.
  2. Tuning Curve Analysis: They recorded the firing rates of individual neurons in the primary motor cortex (M1) during these tasks. They then analyzed how these firing rates varied with the direction of movement, creating "tuning curves."
  3. Hypothesis Testing: To understand the mechanisms behind observed changes in tuning, they tested three hypotheses:
  4. Re-aiming: Monkeys might be aiming at slightly different points than the presented targets.
  5. Speed Gain: Monkeys might be intending to move at different speeds in the 2D vs. 3D tasks.
  6. Dynamic Range Adaptation (DRA): Neurons are adjusting their sensitivity to utilize the available firing rate range more effectively for the current task dimensionality.
  7. Time Course Analysis: They also investigated how quickly this adaptation occurred by analyzing neural responses on a trial-by-trial basis during context switches.

Datasets

The study used data collected from two male rhesus macaques performing BCI tasks. The "datasets" in this context are the neural recordings and behavioral performance metrics from these animals across multiple experimental sessions. The paper does not introduce new benchmark datasets but rather uses data from their experimental setup. The key difference in their experimental design compared to previous work is the manipulation of task dimensionality (2D vs. 3D) within a BCI framework, allowing for a direct comparison of neural responses under controlled conditions.

Key Insights and Ideas

  • Motor Cortex is Adaptive: The most significant finding is that motor cortical neurons, like sensory neurons, exhibit dynamic range adaptation. This suggests that adaptive gain control is a fundamental principle across different brain areas and functions.
  • Dimensionality Matters: The change in task dimensionality (from 3D to 2D) is sufficient to trigger this adaptation in motor cortical neurons. This implies that the brain is sensitive to the complexity and range of possible outputs.
  • DRA as the Dominant Mechanism: The observed changes in neural tuning were best explained by dynamic range adaptation, rather than re-aiming or speed changes. This points to a specific neural strategy for optimizing information coding.
  • Canonical Neural Encoding: The findings suggest that dynamic range adaptation might be a "canonical feature of neural encoding," meaning it's a general mechanism used by the nervous system to efficiently process information.

Key Results

  • Increased Tuning Range in 2D: When switching from a 3D to a 2D task, neurons in M1 showed an increase in their directional tuning range. This means they became more sensitive to variations in movement direction within the 2D plane. Conversely, when moving from 3D to 2D, the range of firing rates for identical stimuli increased.
  • Dose-Response Effect: The degree of tuning change correlated with the "elevation angle" of a neuron's preferred direction in 3D space, a key prediction of the DRA hypothesis. Neurons with preferred directions further from the 2D plane showed larger adaptations.
  • Rapid Adaptation: The adaptation process was remarkably fast, occurring within seconds (estimated at ~1-2 seconds per trial) after the switch in task dimensionality.
  • Hand-Control vs. BCI: Crucially, this dynamic range adaptation was observed in the BCI task but not in a similar task performed with hand control. This highlights the unique advantages of the BCI paradigm for isolating and studying specific neural mechanisms.

Limitations and Future Work

  • Completeness of Adaptation: While adaptation was significant, it wasn't always complete (reaching a theoretical maximum of 1). The authors note that even in sensory systems, adaptation isn't always 100% complete.
  • Hand-Control vs. BCI Differences: The lack of adaptation in the hand-control task is a key limitation. The authors speculate this could be due to the distributed nature of motor control in natural movements, the larger number of parameters involved, or differences in the "control schemes" invoked by physical movement versus BCI control. Future work could explore these differences further.
  • Segregation of Parameters: The paper focuses on directional tuning. Future research could investigate if DRA applies similarly to other movement parameters like speed or force, and whether these adaptations are segregated by parameter category.
  • Functional Implications: While the study demonstrates the neural mechanism, further work is needed to fully understand how these changes in neural sensitivity translate to actual motor behavior and performance.

Summary updated

Abstract

Neural populations from various sensory regions demonstrate dynamic range adaptation in response to changes in the statistical distribution of their input stimuli. These adaptations help optimize the transmission of information about sensory inputs. Here, we show a similar effect in the firing rates of primary motor cortical cells. We trained monkeys to operate a brain-computer interface in both two- and three-dimensional virtual environments. We found that neurons in primary motor cortex exhibited a change in the amplitude of their directional tuning curves between the two tasks. We then leveraged the simultaneous nature of the recordings to test several hypotheses about the population-based mechanisms driving these changes and found that the results are most consistent with dynamic range adaptation. Our results demonstrate that dynamic range adaptation is neither limited to sensory regions nor to rescaling of monotonic stimulus intensity tuning curves, but may rather represent a canonical feature of neural encoding.
BibTeX
@inproceedings{rasmussen2017dynamic,
  author = {R Rasmussen and Andrew B. Schwartz and Steven M. Chase},
  title = {{Dynamic range adaptation in primary motor cortical populations}},
  booktitle = {eLife},
  year = {2017},
  doi = {10.7554/elife.21409},
}
Figure