Temporal dynamics of short-term neural adaptation across human visual cortex

Amber Marijn Brands, Sasha Devore, Orrin Devinsky, Werner Doyle, Adeen Flinker, Daniel Friedman, Patricia Dugan, Jonathan Winawer, Iris I. A. GroenbioRxiv 2023 · DOI

Abstract

Neural responses in visual cortex adapt to prolonged and repeated stimuli. While adaptation occurs across the visual cortex, it is unclear how adaptation patterns and computational mechanisms differ across the visual hierarchy. Here we characterize two signatures of short-term neural adaptation in time-varying intracranial electroencephalography (iEEG) data collected while participants viewed naturalistic image categories varying in duration and repetition interval. Ventral- and lateral-occipitotemporal cortex exhibit slower and prolonged adaptation to single stimuli and slower recovery from adaptation to repeated stimuli compared to V1-V3. For category-selective electrodes, recovery from adaptation is slower for preferred than non-preferred stimuli. To model neural adaptation we augment our delayed divisive normalization (DN) model by scaling the input strength as a function of stimulus category, enabling the model to accurately predict neural responses across multiple image categories. The model fits suggest that differences in adaptation patterns arise from slower normalization dynamics in higher visual areas interacting with differences in input strength resulting from category selectivity. Our results reveal systematic differences in temporal adaptation of neural population responses across the human visual hierarchy and show that a single computational model of history-dependent normalization dynamics, fit with area-specific parameters, accounts for these differences.

Introduction

When the brain encounters a static or repeated stimulus, its neural response typically diminishes over time—a phenomenon known as neural adaptation. This process is thought to enhance coding efficiency by prioritizing novel information. However, it remains debated whether adaptation is a uniform process across the brain or if it varies according to the functional specialization of different cortical regions.

Previous studies often utilized simple stimuli that do not optimally drive higher-order visual areas. Consequently, it has been difficult to determine if observed differences in adaptation across the visual hierarchy are due to stimulus properties or intrinsic differences in temporal processing. This study aims to characterize the temporal dynamics of adaptation across the human visual hierarchy using complex, naturalistic stimuli and high-resolution direct cortical recordings.

Conceptual diagrams of neural adaptation to a single stimulus (A) and repetition suppression to repeated stimuli (B).

Methods

The researchers employed intracranial electroencephalography (iEEG) to record neural activity from 79 visually responsive electrodes in four human participants. This method provides exceptional temporal and spatial resolution. Electrodes were categorized based on their anatomical location into three main regions: primary visual cortex (V1-V3), ventral-occipitotemporal cortex (VOTC), and lateral-occipitotemporal cortex (LOTC).

Anatomical distribution of visually responsive and category-selective electrodes across the visual hierarchy.

Participants performed a simple fixation task while viewing images from six categories (bodies, buildings, faces, objects, scenes, and scrambled images). The experimental design varied both the duration of single stimuli (17 to 533 ms) and the inter-stimulus interval (ISI) between repeated presentations.

To explain the observed neural dynamics, the authors used an augmented Delayed Divisive Normalization (DN) model. In this framework, a neuron's response is normalized by a "leaky" integration of recent activity, creating a history-dependent response. The model was further improved by adding category-selective scaling, allowing it to account for the specific stimulus preferences of different electrodes.

Results

The analysis revealed a systematic "temporal gradient" across the visual cortex.

1. Hierarchy of Temporal Integration

Lower visual areas (V1-V3) responded rapidly to stimulus onset and decayed quickly after offset. In contrast, higher-level areas (VOTC and LOTC) exhibited significantly slower rise times and more prolonged responses. This suggests that higher areas integrate visual information over longer temporal windows, which may be necessary for the stable recognition of complex objects.

2. Differential Recovery Rates

Recovery from adaptation also followed a hierarchical pattern. At the longest repetition interval tested (533 ms), neural responses in V1-V3 had nearly returned to baseline levels. However, responses in VOTC and LOTC remained significantly suppressed, indicating that higher-order regions maintain a longer "memory" of previous visual inputs.

3. Impact of Category Selectivity

For electrodes showing strong category selectivity (e.g., face-selective areas), adaptation was more pronounced for preferred stimuli. Specifically, these areas showed stronger repetition suppression and slower recovery when viewing their preferred category compared to non-preferred categories.

4. Computational Modeling

The augmented DN model successfully captured the diverse neural time courses observed across the hierarchy. By adjusting only the local time constants and input scaling, a single computational architecture was able to predict responses across different brain regions and stimulus categories. This suggests that a common mechanism—delayed normalization—underlies adaptation throughout the visual system.

Discussion

These results demonstrate that neural adaptation is not a uniform process but is tuned to the functional role of each cortical area. The transition from "fast" dynamics in early visual cortex to "slow" dynamics in higher-order areas reflects a fundamental organizational principle of the human visual system.

The success of the delayed divisive normalization model provides a unified computational framework for understanding how the brain manages temporal information. These findings have implications for artificial intelligence; incorporating similar history-dependent normalization into artificial neural networks could improve their ability to process sequential visual data and better mimic human-like perception.

Summary updated

Abstract

Neural responses in visual cortex adapt to prolonged and repeated stimuli. While adaptation occurs across the visual cortex, it is unclear how adaptation patterns and computational mechanisms differ across the visual hierarchy. Here we characterize two signatures of short-term neural adaptation in time-varying intracranial electroencephalography (iEEG) data collected while participants viewed naturalistic image categories varying in duration and repetition interval. Ventral- and lateral-occipitotemporal cortex exhibit slower and prolonged adaptation to single stimuli and slower recovery from adaptation to repeated stimuli compared to V1-V3. For category-selective electrodes, recovery from adaptation is slower for preferred than non-preferred stimuli. To model neural adaptation we augment our delayed divisive normalization (DN) model by scaling the input strength as a function of stimulus category, enabling the model to accurately predict neural responses across multiple image categories. The model fits suggest that differences in adaptation patterns arise from slower normalization dynamics in higher visual areas interacting with differences in input strength resulting from category selectivity. Our results reveal systematic differences in temporal adaptation of neural population responses across the human visual hierarchy and show that a single computational model of history-dependent normalization dynamics, fit with area-specific parameters, accounts for these differences.
BibTeX
@misc{brands2023temporal,
  author = {Amber Marijn Brands and Sasha Devore and Orrin Devinsky and Werner Doyle and Adeen Flinker and Daniel Friedman and Patricia Dugan and Jonathan Winawer and Iris I. A. Groen},
  title = {{Temporal dynamics of short-term neural adaptation across human visual cortex}},
  year = {2023},
  doi = {10.1101/2023.09.13.557378},
  howpublished = {bioRxiv},
}
Figure