Dynamics of visual receptive fields in the macaque frontal eye field

J. Patrick Mayo, Amie DiTomasso, Marc A. Sommer, Matthew A. SmithJournal of Neurophysiology 2015 · DOI

TL;DR

This paper introduces a novel, high-resolution probabilistic method to map the receptive fields (RFs) of neurons in the macaque frontal eye field (FEF). They discovered that FEF neurons exhibit surprisingly precise and rapid temporal integration, responding to stimuli as brief as 10 milliseconds, challenging previous assumptions about their role in visual processing.

What Problem Does This Paper Address?

Understanding how neurons process visual information is fundamental to neuroscience. While receptive fields (RFs) in early visual areas have been extensively mapped using sophisticated techniques, RFs in higher-order areas like the frontal eye field (FEF) remain less understood. Traditional methods for mapping FEF RFs often use limited spatial and temporal parameters, making it difficult to grasp their precise role in complex cognitive functions like attention and motor planning. This lack of detailed understanding hinders our ability to interpret phenomena like presaccadic remapping and the dynamic nature of visual processing. The authors aim to address this gap by developing and applying a more detailed and temporally precise RF mapping technique to FEF neurons.

Methods

The researchers employed a generalized linear model (GLM), a powerful statistical tool, to map the RFs of FEF neurons. This approach is a significant advancement over conventional methods. Instead of presenting simple stimuli at specific locations, they used sparse, full-field, stochastic dot stimuli presented rapidly. The GLM then models the relationship between these complex visual inputs and the neuron's spiking activity.

Key methodological innovations include:

  • Probabilistic RF Mapping: This technique allows for a much richer characterization of RFs, capturing not just the center but also the extent, substructure, and temporal dynamics.
  • High Temporal Resolution: By presenting stimuli in brief frames (as short as 10 ms) and analyzing the temporal kernel of the neuronal response, they could precisely measure temporal integration windows.
  • Passive Fixation Task: This design minimized the influence of motor planning on visual responses, allowing for a clearer assessment of visual processing itself.
  • Validation Against Conventional Methods: The probabilistic RFs (glmRFs) were directly compared to RFs obtained using traditional methods (e.g., memory-guided saccade tasks), demonstrating strong agreement in spatial localization while revealing superior detail.

Datasets

The study used data collected from two adult rhesus macaques. The visual stimuli were presented on a CRT monitor, and eye position was tracked. No new benchmark datasets were introduced; rather, the study utilized novel stimulus presentation and analysis techniques on data collected from these subjects.

Key Insights and Ideas

  • FEF Neurons are Faster Than We Thought: The most striking finding is the FEF neurons' ability to integrate visual information over remarkably short timescales (as brief as 10 ms). This suggests a more direct and rapid role in visual processing than previously appreciated.
  • Probabilistic Mapping Unlocks Detail: The GLM-based probabilistic approach revealed finer details of FEF RFs, including their spatial extent and temporal dynamics, which were not apparent with conventional mapping.
  • Spatial Structure is Stable: Despite the dynamic nature of visual processing, the spatial structure of FEF RFs remained largely consistent over time and across different stimulus conditions.
  • FEF as a Visual Area: The results reinforce the idea that FEF is not solely a motor planning area but also a crucial component of the visual processing stream, receiving and processing visual inputs with high temporal precision.

Key Results

  • Precise Temporal Integration: FEF neurons showed robust responses to stimuli presented as briefly as 10 ms, and their temporal integration windows were found to be relatively short (average optimal window duration of 118 ms).
  • High Agreement with Conventional RFs: The spatial locations and eccentricities of the glmRFs closely matched those obtained through conventional mapping, validating the new approach.
  • Detailed RF Characterization: The probabilistic mapping provided significantly more detailed RF maps compared to conventional methods, revealing substructure and extent.
  • No Bias in Mapping: The study found no significant correlation between neuronal firing characteristics (baseline rate, VMI, TSI, SNR) and the goodness of fit of the glmRFs, suggesting the mapping method is unbiased.
  • Rapid Data Acquisition: Acquiring a high-quality glmRF was significantly faster at higher stimulus presentation rates (e.g., 30 ms per frame), indicating efficiency gains with the new method.

Limitations and Future Work

  • Passive Fixation: While useful for isolating visual responses, the passive fixation task doesn't fully capture the complex, active vision scenarios FEF neurons operate in during natural behavior. The authors suggest that future work should expand to more detailed experimental tasks.
  • Spatial Interactions: The study acknowledges that spatial interactions (e.g., suppressive surrounds) were observed in a few neurons but were not extensively explored due to the small sample size for certain tasks. They propose that their method could be adapted to investigate these interactions further.
  • Model Complexity: While the GLM with L1-norm regularization was effective, the authors note that more complex models (e.g., mixture of Gaussians) might yield even better fits for RFs.
  • Future Directions: The authors highlight the potential of their method for studying RF dynamics during perisaccadic vision, understanding attention-driven RF shifts, and exploring the neural circuitry underlying visual stability and rapid cognitive processing. They also suggest that computational advances will make such detailed analyses more feasible.

Summary updated

Abstract

Neuronal receptive fields (RFs) provide the foundation for understanding systems-level sensory processing. In early visual areas, investigators have mapped RFs in detail using stochastic stimuli and sophisticated analytical approaches. Much less is known about RFs in prefrontal cortex. Visual stimuli used for mapping RFs in prefrontal cortex tend to cover a small range of spatial and temporal parameters, making it difficult to understand their role in visual processing. To address these shortcomings, we implemented a generalized linear model to measure the RFs of neurons in the macaque frontal eye field (FEF) in response to sparse, full-field stimuli. Our high-resolution, probabilistic approach tracked the evolution of RFs during passive fixation, and we validated our results against conventional measures. We found that FEF neurons exhibited a surprising level of sensitivity to stimuli presented as briefly as 10 ms or to multiple dots presented simultaneously, suggesting that FEF visual responses are more precise than previously appreciated. FEF RF spatial structures were largely maintained over time and between stimulus conditions. Our results demonstrate that the application of probabilistic RF mapping to FEF and similar association areas is an important tool for clarifying the neuronal mechanisms of cognition.
BibTeX
@article{mayo2015dynamics,
  author = {J. Patrick Mayo and Amie DiTomasso and Marc A. Sommer and Matthew A. Smith},
  title = {{Dynamics of visual receptive fields in the macaque frontal eye field}},
  journal = {Journal of Neurophysiology},
  year = {2015},
  doi = {10.1152/jn.00746.2015},
}

Loading related content...

Figure