Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Functional embeddings enable Aggregation of multi-area SEEG recordings over subjects and sessions

Sina Javadzadeh, Rahil Soroushmojdehi, S. Alireza Seyyed Mousavi, Mehrnaz Asadi, Sumiko Abe, Terence D. Sanger

TL;DR

A scalable representation-learning framework that learns a subject-agnostic functional identity for each electrode from multi-region local field potentials using a Siamese encoder with contrastive objectives, inducing an embedding geometry that is locality-sensitive to region-specific neural signatures, and tokenizes these embeddings for a transformer that models inter-regional relationships with a variable number of channels.

Abstract

Aggregating intracranial recordings across subjects is challenging since electrode count, placement, and covered regions vary widely. Spatial normalization methods like MNI coordinates offer a shared anatomical reference, but often fail to capture true functional similarity, particularly when localization is imprecise; even at matched anatomical coordinates, the targeted brain region and underlying neural dynamics can differ substantially between individuals. We propose a scalable representation-learning framework that (i) learns a subject-agnostic functional identity for each electrode from multi-region local field potentials using a Siamese encoder with contrastive objectives, inducing an embedding geometry that is locality-sensitive to region-specific neural signatures, and (ii) tokenizes these embeddings for a transformer that models inter-regional relationships with a variable number of channels. We evaluate this framework on a 20-subject dataset spanning basal ganglia-thalamic regions collected during flexible rest/movement recording sessions with heterogeneous electrode layouts. The learned functional space supports accurate within-subject discrimination and forms clear, region-consistent clusters; it transfers zero-shot to unseen channels. The transformer, operating on functional tokens without subject-specific heads or supervision, captures cross-region dependencies and enables reconstruction of masked channels, providing a subject-agnostic backbone for downstream decoding. Together, these results indicate a path toward large-scale, cross-subject aggregation and pretraining for intracranial neural data where strict task structure and uniform sensor placement are unavailable.

Functional embeddings enable Aggregation of multi-area SEEG recordings over subjects and sessions

TL;DR

A scalable representation-learning framework that learns a subject-agnostic functional identity for each electrode from multi-region local field potentials using a Siamese encoder with contrastive objectives, inducing an embedding geometry that is locality-sensitive to region-specific neural signatures, and tokenizes these embeddings for a transformer that models inter-regional relationships with a variable number of channels.

Abstract

Aggregating intracranial recordings across subjects is challenging since electrode count, placement, and covered regions vary widely. Spatial normalization methods like MNI coordinates offer a shared anatomical reference, but often fail to capture true functional similarity, particularly when localization is imprecise; even at matched anatomical coordinates, the targeted brain region and underlying neural dynamics can differ substantially between individuals. We propose a scalable representation-learning framework that (i) learns a subject-agnostic functional identity for each electrode from multi-region local field potentials using a Siamese encoder with contrastive objectives, inducing an embedding geometry that is locality-sensitive to region-specific neural signatures, and (ii) tokenizes these embeddings for a transformer that models inter-regional relationships with a variable number of channels. We evaluate this framework on a 20-subject dataset spanning basal ganglia-thalamic regions collected during flexible rest/movement recording sessions with heterogeneous electrode layouts. The learned functional space supports accurate within-subject discrimination and forms clear, region-consistent clusters; it transfers zero-shot to unseen channels. The transformer, operating on functional tokens without subject-specific heads or supervision, captures cross-region dependencies and enables reconstruction of masked channels, providing a subject-agnostic backbone for downstream decoding. Together, these results indicate a path toward large-scale, cross-subject aggregation and pretraining for intracranial neural data where strict task structure and uniform sensor placement are unavailable.

Paper Structure

This paper contains 13 sections, 5 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Transformers for neural population modeling
  4. Neural decoding with SEEG
  5. Cross-subject SEEG decoding under electrode variability
  6. Method
  7. Data Collection and Preprocessing
  8. Functional Embedding Network
  9. Details.
  10. Functional Transformer: Masked-Region Reconstruction
  11. Results
  12. Functional embeddings reveal neural signature of each brain region across subjects (simulation study)
  13. Functional Embedding learned within individual subjects provide a locality sensitive embedding space encoding function

Figures (2)

  • Figure 1: Overview of Approach. 1: Contrastive learning framework used to train the CNN-based functional encoder via a Siamese architecture, encouraging signals from the same brain region to map nearby in latent space. 2: Learned functional embeddings produced by the trained encoder, showing clustering by brain region. 3: Incorporating functional embeddings into the tokenization scheme. 4: Functional Transformer architecture that enables scalable modeling across subjects and sessions .
  • Figure 2: Simulation-based validation of the functional embedding.(A) Contrastive encoder trained on simulated data clusters test segments by region and generalizes to held-out subjects (scatter, top) with high accuracy (confusion matrix, bottom). (B) Top: Power spectra of signals nearest each region centroid match the ground-truth signatures; Bottom: Perturbation saliency shows largest embedding shifts when disrupting signature bursts. (C) Smooth frequency sweeps trace continuous trajectories in the embedding, passing near region centroids tuned to corresponding bands, indicating a locality-sensitive functional map.