STAMP-LME: A spatiotemporal surface-mapping framework for population-level iEEG

Intracranial EEG (iEEG) offers unparalleled advantages in capturing human neural activity with high temporal and spatial resolution. However, conducting group-level statistical analyses in iEEG studies proves challenging due to limited electrode coverage, anatomical differences among individuals, and the complexity of integrating spatiotemporal responses across participants.

Researchers often rely on predefined brain regions of interest, which compromises the spatial resolution of iEEG and eliminates data points outside these regions. We introduce STAMP-LME, a spatiotemporal surface-mapping framework that enables comprehensive population-level iEEG analysis without relying on predefined cortical regions of interest.

The STAMP-LME pipeline encompasses electrode localization, mapping to a standardized cortical surface, and statistical inference at the cortical level.

Each participant's electrode-level data is meticulously mapped to a standardized cortical surface using weights that consider electrode-to-cortex distance and confidence in cortical sampling. Brain signals are then smoothed spatially and entered into linear mixed-effects models (LME) at each vertex-node and time point. Data-derived null distributions are generated through controlled permutations, facilitating spatiotemporal inference using threshold-free cluster enhancement (TFCE) or false-discovery-rate (FDR) correction.

To validate the efficacy of STAMP-LME, we conducted simulations with known cortical ground truth while varying effect sizes, interparticipant source-location variability, electrode coverage, and spatial smoothing parameters. The framework demonstrated robust detection of simulated effects while maintaining accurate localization and minimizing false-positive detections across various sampling conditions.

We then applied STAMP-LME to intracranial recordings from nine individuals performing an auditory oddball task. The empirical analysis yielded distributed auditory cortical responses that align with conventional channel-level analyses, while providing more precise group-level activation maps across sampled cortical regions. STAMP-LME offers a flexible, open-source solution for data-driven population-level cortical iEEG analysis, preserving the spatiotemporal benefits of intracranial recordings while addressing the challenges posed by sparse and heterogeneous electrode coverage.

This structured workflow enables reproducible and anatomically precise group-level inference on a standardized cortical surface, facilitating comparisons with surface-based fMRI and source-localized M/EEG results. By providing a practical tool for cognitive and clinical neuroscience research, STAMP-LME significantly advances the field of iEEG analysis.