Neural population dynamics in human motor cortex during movements in people with ALS

Chethan Pandarinath, Vikash Gilja, Christine H Blabe, Paul Nuyujukian, Anish A. Sarma, Brittany L Sorice, Emad N. Eskandar, Leigh R. Hochberg, Jaimie M. Henderson, Krishna V. ShenoyeLife 2015 · DOI

TL;DR

This study reveals that human motor cortex activity during movement exhibits a "rotational" dynamical structure, similar to what's observed in non-human primates. This suggests a fundamental, shared mechanism for movement generation across species, even in individuals with severe motor impairment due to ALS.

What Problem Does This Paper Address?

For decades, the prevailing view of the motor cortex has been that individual neurons directly encode specific movement parameters like direction, speed, or force. However, recent research in non-human primates has challenged this "direct messaging" hypothesis. Instead, studies suggest that motor cortex activity is better described by a dynamical system, where the collective activity of neuronal populations evolves over time in a predictable, "rotational" pattern. This pattern isn't directly tied to specific movement features but rather represents an underlying computational structure for generating movement. The key question this paper addresses is whether this dynamical system perspective also applies to human motor cortex, particularly in the context of movement disorders like Amyotrophic Lateral Sclerosis (ALS), where motor control is significantly impaired.

Methods

The researchers employed a sophisticated analytical technique called jPCA (joint Principal Component Analysis). This method is designed to identify and visualize rotational patterns within high-dimensional neural population data. Here's a breakdown of their approach:

  1. Data Acquisition: They recorded neural activity (action potentials or "spikes") from the motor cortex of two participants with ALS using implanted microelectrode arrays.
  2. Task Design: Participants performed a "center-out" target acquisition task using a touchpad to control a cursor. This involved moving the cursor from a central point to one of eight peripheral targets.
  3. Dimensionality Reduction: Before applying jPCA, they used standard Principal Component Analysis (PCA) to reduce the dimensionality of the neural data, focusing on the most prominent response patterns (the top 6 principal components). This step helps to isolate the core dynamics from noise.
  4. jPCA Application: jPCA was then applied to this reduced-dimensional data. The core idea of jPCA is to find a 2D plane within the neural state space where the activity exhibits the strongest rotational tendencies over time. This plane is identified by finding the eigenvectors of a specific matrix that captures the neural state's evolution.
  5. Control Analyses: To ensure that the observed rotations were not simply an artifact of the jPCA method or complex neural patterns, they performed rigorous shuffle control analyses. These involved disrupting the temporal structure of the neural data while preserving individual neuron response complexity. If rotations persisted after shuffling, it would suggest a spurious finding.

Datasets

The study utilized neural recordings from two human participants (T6 and T7) diagnosed with Amyotrophic Lateral Sclerosis (ALS). These participants were enrolled in the BrainGate2 pilot clinical trial. The data were collected during a "center-out" target acquisition task. This is not a newly introduced benchmark dataset but rather novel data collected specifically for this research from human subjects with a motor impairment.

Key Insights and Ideas

  • Human Motor Cortex Operates Dynamically: The most significant insight is that human motor cortex, like primate motor cortex, exhibits a fundamental "rotational" dynamical structure during movement. This challenges the simpler view of direct parameter encoding.
  • Universality of Motor Control Mechanisms: The similarity in dynamics between humans (even those with ALS) and non-human primates suggests that the underlying computational principles of motor control are conserved across species.
  • ALS Doesn't Erase Core Dynamics: Despite significant motor impairment, the participants' motor cortex still displayed these complex rotational patterns. This implies that the core dynamical system for movement generation might be remarkably resilient to severe neurodegenerative disease.
  • Implications for Brain-Machine Interfaces (BMIs): Understanding these dynamics is crucial for improving BMI technology. Incorporating knowledge of these population-level dynamics could lead to more fluid and intuitive control of prosthetic devices.
  • Questioning the Prevailing Model: The findings strongly support a dynamical systems perspective over the traditional parameter-encoding model for motor cortical activity in humans.

Key Results

  • Prominent Rotational Dynamics: jPCA analysis revealed strong rotational dynamics in the neural population activity for both participants (T6 and T7).
  • Consistency Across Conditions: These rotations proceeded in the same direction across different movement conditions (i.e., different target locations), indicating a consistent underlying system.
  • Variance Explained: The top jPCA plane captured a substantial portion of the neural data's variance: 61% for participant T6 and 27% for participant T7. For comparison, similar analyses in macaques captured around 28%.
  • Control Analysis Success: Shuffle control analyses demonstrated that the observed rotational structure was not an artifact of the jPCA method. Rotations were significantly diminished or absent in the shuffled data.
  • Angle Consistency: Histograms of the angle between the neural state and its derivative consistently peaked near π/2, a signature of rotational dynamics, similar to findings in macaques.

Limitations and Future Work

  • ALS Specificity: The authors acknowledge that the findings do not conclusively prove the presence of these dynamics in healthy human motor cortex, as both participants had ALS and likely experienced disease-related changes in their motor cortex.
  • Future Directions:
    • Healthy Controls: Further experiments with healthy human participants are needed to confirm these findings in the absence of disease.
  • Imagined Movements: Investigating whether similar dynamical activity exists during imagined movements is a key open question. If dynamics are absent during imagined movements, it could support their role as an "oscillatory basis set" for generating motor output.
  • BMI Improvement: The study highlights the potential for incorporating neural population dynamics into BMI control algorithms to enhance performance.

Summary updated

Abstract

The prevailing view of motor cortex holds that motor cortical neural activity represents muscle or movement parameters. However, recent studies in non-human primates have shown that neural activity does not simply represent muscle or movement parameters; instead, its temporal structure is well-described by a dynamical system where activity during movement evolves lawfully from an initial pre-movement state. In this study, we analyze neuronal ensemble activity in motor cortex in two clinical trial participants diagnosed with Amyotrophic Lateral Sclerosis (ALS). We find that activity in human motor cortex has similar dynamical structure to that of non-human primates, indicating that human motor cortex contains a similar underlying dynamical system for movement generation.
BibTeX
@inproceedings{pandarinath2015neural,
  author = {Chethan Pandarinath and Vikash Gilja and Christine H Blabe and Paul Nuyujukian and Anish A. Sarma and Brittany L Sorice and Emad N. Eskandar and Leigh R. Hochberg and Jaimie M. Henderson and Krishna V. Shenoy},
  title = {{Neural population dynamics in human motor cortex during movements in people with ALS}},
  booktitle = {eLife},
  year = {2015},
  doi = {10.7554/elife.07436},
}
Figure