Cortical Topography of Error-Related High-Frequency Potentials During Erroneous Control in a Continuous Control Brain–Computer Interface

Nile Rebecca Wilson, Devapratim Sarma, Jeremiah Wander, Kurt E. Weaver, Jeffrey G. Ojemann, Rajesh P. N. RaoFrontiers in Neuroscience 2019 · DOI

TL;DR

This study investigates how the brain responds to errors in a continuous control brain-computer interface (BCI) using electrocorticography (ECoG). They found that specific high-frequency brain activity patterns (high-gamma band power) emerge in various cortical areas within 100-500 ms after a BCI makes a decoding error, offering a potential for automatic, task-independent feedback to improve BCI performance.

What Problem Does This Paper Address?

Current brain-computer interfaces (BCIs) often rely on explicit user feedback or task-specific information to adapt and improve. However, there's a need for automatic, task-independent feedback mechanisms that can enhance BCI usability and robustness. Error-related potentials (ErrPs), brain responses generated when a user makes a mistake, have shown promise as such feedback signals. While previous research has explored ErrPs, most studies have focused on discrete BCI tasks or used less invasive EEG, which has limitations in spatial resolution. This paper addresses the gap of understanding how ErrPs manifest in a more naturalistic, continuous control BCI setting using high-resolution ECoG, specifically focusing on the topographical distribution of high-frequency brain activity.

Methods

The researchers employed electrocorticography (ECoG) to record brain activity from three human subjects with medically intractable epilepsy. These subjects performed a one-dimensional center-out BCI task where they controlled a cursor's velocity using either motor imagery or rest. The BCI decoder's performance was monitored, and epochs where the decoder incorrectly moved the cursor opposite to the subject's intended direction were identified as "error epochs." Conversely, epochs where the cursor moved as intended were labeled "correct epochs."

The core analysis focused on high-gamma band power (70-100 Hz), which is known to reflect local neural activity. They extracted statistical analysis epochs from 100-500 ms after the identified error or correct movement onset. Statistical analyses, including two-way ANOVAs and post-hoc t-tests, were used to compare high-gamma band power between error and correct conditions across different Brodmann Areas (BAs) of the cortex. This allowed them to map the "topography" of these error-related brain responses.

Datasets

The study utilized data collected from three human subjects undergoing clinical seizure monitoring. These subjects participated in a custom-designed BCI task. No pre-existing datasets were used for training or evaluation. The data collected from these three participants served as the sole dataset for the study.

Key Insights and Ideas

  • Error Signals are Topographically Specific: The study reveals that errors in a continuous control BCI elicit distinct patterns of high-gamma band power across various cortical regions, not just a single localized response.
  • High-Gamma Band Power as a Reliable Indicator: High-gamma band power (70-100 Hz) proved to be a sensitive measure for detecting error-related neural activity, offering better spatial localization than lower frequency bands.
  • Motor Imagery vs. Rest Control Differences: While many cortical areas showed increased high-gamma power during errors regardless of whether the subject used motor imagery or rest, some areas showed exclusive responses depending on the control modality. This suggests that the brain's error processing might be modulated by the specific motor control strategy employed.
  • Continuous Control Offers Ecological Validity: By focusing on continuous control, the study provides a more naturalistic and ecologically valid setting for understanding error processing compared to discrete BCI tasks. This is crucial for developing practical, real-world BCI applications.
  • Potential for Task-Independent Feedback: The identified error-related high-gamma responses could serve as a valuable source of task-independent feedback for BCI decoders, enabling more robust and personalized adaptive BCI systems.

Key Results

  • Increased high-gamma band power (HBP) was observed in primary somatosensory, motor, premotor, and parietal areas within 100-500 ms following BCI errors, compared to correct control epochs.
  • These responses were generally observed regardless of whether the subject was using motor imagery or rest to control the cursor.
  • Specific parts of the somatosensory, temporal, and parietal areas showed increased HBP exclusively during motor imagery error.
  • A portion of the parietal cortex near the angular gyrus exclusively showed increased HBP during rest error.
  • The study found statistically significant interactions between trial type (motor imagery vs. rest) and performance (error vs. correct) in several Brodmann Areas, including BA 3, BA 4, BA 40, and BA 43.

Limitations and Future Work

  • Limited Number of Subjects: The study involved only three subjects, which limits the generalizability of the findings. Future research with a larger cohort is needed.
  • Low Task Performance: The subjects exhibited relatively low overall trial success rates. While the authors argue this still validates the findings (as error potentials can be elicited even when errors are frequent), it might influence the nature of the error signals.
  • Post Hoc Error Identification: The error epochs were identified post hoc based on behavioral mismatches. While this provides ecological validity, it means the error onset is not as precisely time-locked as in experimentally induced error paradigms, potentially leading to less distinct waveforms.
  • Future Work: The authors suggest exploring the long-term effects of learning on these error-related potentials and developing robust classification techniques for real-time error detection. They also propose investigating other forms of feedback beyond visual feedback for continuous control BCIs. The potential for using these findings to develop co-adaptive BCIs that automatically adjust based on error signals is a key future direction.

Summary updated

Abstract

Brain-computer interfaces (BCIs) benefit greatly from performance feedback, but current systems lack automatic, task-independent feedback. Cortical responses elicited from user error have the potential to serve as state-based feedback to BCI decoders. To gain a better understanding of local error potentials, we investigate responsive cortical power underlying error-related potentials (ErrPs) from the human cortex during a one-dimensional center-out BCI task, tracking the topography of high-gamma (70-100 Hz) band power (HBP) specific to BCI error. We measured electrocorticography (ECoG) in three human subjects during dynamic, continuous control over BCI cursor velocity. Subjects used motor imagery and rest to move the cursor toward and subsequently dwell within a target region. We then identified and labeled epochs where the BCI decoder incorrectly moved the cursor in the direction opposite of the subject's expectations (i.e., BCI error). We found increased HBP in various cortical areas 100-500 ms following BCI error with respect to epochs of correct, intended control. Significant responses were noted in primary somatosensory, motor, premotor, and parietal areas and generally regardless of whether the subject was using motor imagery or rest to move the cursor toward the target. Parts of somatosensory, temporal, and parietal areas exclusively had increased HBP when subjects were using motor imagery. In contrast, only part of the parietal cortex near the angular gyrus exclusively had an increase in HBP during rest. This investigation is, to our knowledge, the first to explore cortical fields changes in the context of continuous control in ECoG BCI. We present topographical changes in HBP characteristic specific to the generation of error. By focusing on continuous control, instead of on discrete control for simple selection, we investigate a more naturalistic setting and provide high ecological validity for characterizing error potentials. Such potentials could be considered as design elements for co-adaptive BCIs in the future as task-independent feedback to the decoder, allowing for more robust and individualized BCIs.
BibTeX
@inproceedings{wilson2019cortical,
  author = {Nile Rebecca Wilson and Devapratim Sarma and Jeremiah Wander and Kurt E. Weaver and Jeffrey G. Ojemann and Rajesh P. N. Rao},
  title = {{Cortical Topography of Error-Related High-Frequency Potentials During Erroneous Control in a Continuous Control Brain–Computer Interface}},
  booktitle = {Frontiers in Neuroscience},
  year = {2019},
  doi = {10.3389/fnins.2019.00502},
}
Figure