Epidural field potentials (EFPs), recorded using epidural microelectrode arrays, offer a less invasive alternative to other methods for recording brain activity. While promising for both basic and clinical neuroscience, the precise information content of EFPs remains unclear. This study leverages the well-understood retinotopic and columnar organization of macaque V1 to investigate the extent to which EFPs preserve information about various visual stimulus features, including location, size, shape, and color. These features were chosen to probe different spatial scales of functional organization within V1. The hypothesis is that despite the integration of signals from numerous neurons, EFPs will retain sufficient detail to allow for reliable discrimination of stimuli based on these features. This research aims to bridge the gap in understanding EFP information content, potentially expanding their use in chronic recordings for brain-computer interfaces (BCIs) and other applications. Previous work has shown promising results in decoding spatial information from V1 EFPs, but the contribution of non-spatial features remained unclear. This study addresses this by systematically manipulating multiple stimulus parameters, enabling a more complete understanding of EFP selectivity and its potential for clinical and research applications. The high spatial resolution and columnar structure of V1 provides a beneficial context for this investigation, allowing for the assessment of information processing at different spatial scales. This work directly addresses the limitations of current knowledge about EFPs by disentangling the contributions of different visual features to the overall signal, potentially improving upon the limited understanding that constrains the use of epidural recordings over more invasive approaches.

Several studies have explored the potential of EFPs, highlighting their advantages in terms of reduced invasiveness compared to intracortical recordings. However, some studies suggest that EFPs may have lower specificity due to the distance between electrodes and neurons, leading to larger cortical signal spread and potentially higher noise levels. Conversely, other research has reported comparable decoding performance between EFPs and local field potentials (LFPs), indicating that the dura mater does not significantly impede feature detection. These conflicting results highlight the need for further investigation into the information content of EFPs. Recent findings demonstrating high spatial information decoding from short fragments of V1 EFPs emphasize their potential for clinical and BCI applications. However, these studies often conflate the contributions of multiple stimulus features, limiting a clear understanding of EFP selectivity. The current study seeks to address this knowledge gap by focusing on the specific contributions of location, size, shape, and color to the EFP signal in V1.

Two male macaque monkeys were implanted with epidural microelectrode arrays over the left V1 hemisphere. Electrode arrays consisted of high-amplitude electrodes with 1.8 mm center-to-center distances. EFPs were recorded at 25 kHz and referenced against a larger electrode in the skull. Visual stimuli comprised various shapes (triangles, squares, circle) in different sizes (1°, 2°, 14° diameter) and colors (four chromatic plus achromatic) presented at five locations in the visual field. The stimuli were presented sequentially in a pseudorandom order. The monkeys performed a fixation task where they were rewarded for detecting a dimming of the fixation point, ensuring attention to the stimuli. The recorded EFPs were preprocessed using a low-pass filter, downsampling, and wavelet transformation. Receptive fields (RFs) for each electrode were determined using an automated bar-mapping procedure based on reverse correlation. The most informative time-frequency ranges for analyzing EFP modulation for different features were determined using a receiver-operating characteristic (ROC)-based approach. Stimulus-specific EFP modulations were analyzed for each feature category. Spatial sensitivity was assessed by comparing responses to stimuli at different locations relative to the electrode's RF. Size, shape, and color sensitivities were quantified by comparing responses across different stimulus sizes, shapes, and colors, respectively. Single-trial classification performance was evaluated using support vector machines (SVMs) to determine the extent to which non-spatial features (size, shape, color) could be decoded from single trials.

The study found that EFPs in macaque V1 provide detailed information about visual stimuli. EFP responses were highly sensitive to stimulus location, with modulation extending within 1.5 degrees of the RF center, exceeding the average RF size (∼3 degrees). Significant modulation was also observed for differences in stimulus size (as small as 0.2-degree diameter differences), shape (based on orientation statistics rather than global orientation), and color (with a bias toward blue and red stimuli evident during later response periods). The most informative time-frequency ranges varied across features: high gamma (≥85 Hz) for location, mid-gamma (30–80 Hz) for size and shape, and distinct gamma ranges for color. Remarkably, single-trial classification using non-spatial features (size, shape, color) achieved high accuracy (75.6–85.5%), demonstrating the reliability and specificity of EFP modulation. These findings were consistent across the two monkeys in the study.

The results demonstrate the high sensitivity and functional specificity of V1 EFPs, suggesting they offer a powerful yet minimally invasive approach to recording brain activity. The observed modulation of EFPs by subtle differences in non-spatial stimulus features highlights the remarkable information content of these signals, even when considering the integration of activity from numerous neurons. The ability to achieve high single-trial classification accuracy using non-spatial features further underscores the utility of EFPs in applications such as brain-computer interfaces. The findings challenge previous assumptions about EFP limitations, suggesting that they can provide a valuable alternative to more invasive recording techniques. Future research should investigate the extent to which these findings generalize to other cortical areas and the potential for improving EFP recording techniques to further enhance signal quality and spatial resolution.

This study provides compelling evidence that EFPs in macaque V1 carry rich information about visual stimuli, including spatial and non-spatial features. The high sensitivity to location and significant modulation by size, shape, and color, along with the demonstrated single-trial classification accuracy, underscores the value of EFPs for both basic research and clinical applications. Future work should focus on refining EFP recording technologies and extending these findings to other cortical regions and sensory modalities.

The study was conducted in macaque monkeys, which may limit the generalizability of the findings to humans. The specific experimental setup and stimulus parameters might also influence the observed results. The relatively small number of electrodes used could also restrict the ability to determine certain feature sensitivities across different regions. Further research is needed to fully understand the impact of these factors on EFP characteristics and their potential applications.