Large-scale dynamics of perceptual decision information across human cortex

Niklas Wilming, Peter R. Murphy, Florent Meyniel, Tobias H. DonnerNature Communications 2020 · DOI

TL;DR

This study uses MEG and advanced decoding techniques to map how visual information is processed and integrated across the human brain during perceptual decision-making. They discovered that early visual cortex (like V1) not only processes immediate sensory input but also contains an "endogenous" signal that contributes to decisions, communicating with motor areas via low-frequency brain waves.

What Problem Does This Paper Address?

Perceptual decisions, like deciding if a visual stimulus is brighter than a reference, involve accumulating sensory evidence over time. A dominant theory suggests sensory areas process the raw evidence, which is then passed to downstream areas for accumulation into a decision variable, ultimately leading to an action plan. However, how this entire process unfolds across the vast network of the human cortex, and whether there are internal signals beyond direct sensory input, has been unclear. Specifically, the role of feedback connections from higher-order areas back to sensory cortex in shaping these decisions, and how stimulus-dependent versus internally generated signals contribute, remained largely unexplored at a large scale.

Methods

The researchers employed a sophisticated combination of techniques:

  • Magnetoencephalography (MEG): This non-invasive neuroimaging method measures magnetic fields produced by electrical activity in the brain, offering excellent temporal resolution.
  • Regionally-Specific Decoding: They developed an approach to decode specific information (like stimulus contrast or the chosen action) from MEG signals within defined cortical regions of interest (ROIs). This involved using machine learning classifiers trained on the spectro-spatial patterns of brain activity.
  • Behavioral Analysis: Participants performed a visual task where they had to decide if a test stimulus's contrast was higher or lower than a reference. Their choices and confidence were recorded, and their behavior was analyzed to infer the time course of evidence accumulation using psychophysical kernels.
  • Frequency-Specific Analysis: The study specifically analyzed brain activity in different frequency bands (e.g., gamma band for high frequencies, alpha/beta bands for low frequencies) to disentangle the roles of different neural processes.
  • Disentangling Stimulus-Dependent vs. Endogenous Signals: By analyzing "residual kernels" (activity remaining after accounting for the direct influence of the stimulus), they could identify internally generated signals contributing to the decision.

Datasets

The study used data collected from 15 human participants performing a visual decision-making task. No new benchmark datasets were introduced; rather, existing fMRI-defined ROIs were used to guide the analysis of MEG data.

Key Insights and Ideas

  • Early Visual Cortex Isn't Just a Passive Receiver: Contrary to a purely feedforward model, the study found evidence for an "endogenous" component of decision-related activity in early visual areas like V1. This suggests these areas are not just processing incoming sensory data but also contributing internal signals to the decision process.
  • Frequency Bands Carry Different Information: The research highlights a spectral multiplexing of information. Gamma-band activity (high frequency) in visual cortex seems to reflect the immediate sensory evidence, while low-frequency activity (specifically alpha/beta bands) carries an endogenous signal that contributes to the decision.
  • Feedback Loops are Crucial: The low-frequency endogenous signal in early visual cortex appears to track the build-up of choice-predictive activity in downstream motor and premotor regions, with a noticeable delay. This strongly suggests a feedback mechanism where higher-level decision signals influence earlier sensory processing.
  • A Hierarchical, Dynamic Process: The findings paint a picture of perceptual decision-making as a dynamic, large-scale cortical process. Sensory evidence is processed, but internal signals and feedback loops continuously shape the evolving decision variable across multiple brain regions.

Key Results

  • Stimulus-Dependent Signal: Gamma-band activity in visual cortex (e.g., V1) reflected the instantaneous sensory input and decayed over time, consistent with adaptation. This signal was strongly correlated with behavioral evidence weighting.
  • Endogenous Signal: Low-frequency (0-20 Hz) activity in V1 showed choice-predictive fluctuations that were not explained by the external stimulus. This endogenous signal lagged behind the build-up of choice-predictive activity in motor and parietal areas by approximately 150 ms.
  • Cortical Pathway: Sensory areas (visual cortex) were primarily involved in encoding sensory evidence, while association and motor areas (parietal, premotor, motor cortex) showed the build-up of choice-predictive activity, reflecting evidence accumulation and action planning.
  • Frequency-Specific Interactions: A negative interaction was observed between low-frequency and gamma-band activity in visual cortex, suggesting that the endogenous low-frequency signal might modulate the influence of sensory evidence (gamma-band) on the final choice, particularly later in the decision process.

Limitations and Future Work

  • MEG Limitations: While MEG offers excellent temporal resolution, its spatial resolution is inherently limited. The study acknowledges that signal leakage between ROIs can be a concern, although they employed methods to mitigate this.
  • Causality: The study demonstrates correlations and associations between brain activity and behavior. Establishing causal links would require interventional studies (e.g., using TMS or optogenetics).
  • Generalizability: The findings are based on a specific visual decision task. Future work could explore whether these principles apply to other sensory modalities or more complex decision scenarios.
  • Nature of Endogenous Signal: While identified, the precise neural mechanisms generating and propagating this endogenous signal require further investigation.
  • Feedback Pathways: The study suggests feedback signaling but doesn't pinpoint the exact anatomical pathways involved. Future research could focus on mapping these specific feedback connections and their role in decision formation.

Summary updated

Abstract

Perceptual decisions entail the accumulation of sensory evidence for a particular choice towards an action plan. An influential framework holds that sensory cortical areas encode the instantaneous sensory evidence and downstream, action-related regions accumulate this evidence. The large-scale distribution of this computation across the cerebral cortex has remained largely elusive. Here, we develop a regionally-specific magnetoencephalography decoding approach to exhaustively map the dynamics of stimulus- and choice-specific signals across the human cortical surface during a visual decision. Comparison with the evidence accumulation dynamics inferred from behavior disentangles stimulus-dependent and endogenous components of choice-predictive activity across the visual cortical hierarchy. We find such an endogenous component in early visual cortex (including V1), which is expressed in a low (<20 Hz) frequency band and tracks, with delay, the build-up of choice-predictive activity in (pre-) motor regions. Our results are consistent with choice- and frequency-specific cortical feedback signaling during decision formation.
BibTeX
@inproceedings{wilming2020large,
  author = {Niklas Wilming and Peter R. Murphy and Florent Meyniel and Tobias H. Donner},
  title = {{Large-scale dynamics of perceptual decision information across human cortex}},
  booktitle = {Nature Communications},
  year = {2020},
  doi = {10.1038/s41467-020-18826-6},
}
Figure