MtReNIRS: A Self-Supervised Foundation Model for Transferable Neurovascular Representations
Abstract
Functional near-infrared spectroscopy (fNIRS) offers a promising approach for monitoring the hemodynamic correlates of neural activity in naturalistic settings. However, existing modeling methods for fNIRS are largely confined to small-scale, single-session data with low sampling rates, driven by high heterogeneity in acquisition devices and optode montages. In addition, the complex spatiotemporal structure of fNIRS signals further restricts model scalability and transferability across settings. As a result, it remains unclear whether a single fNIRS encoder can learn representations that are usable across subjects, tasks, devices, and montages. Here, we show that a montage-aware masked autoencoder, pretrained on 12 public fNIRS datasets spanning 268 subjects and 567 recording hours, learns generalizable neurovascular representations that transfer across three external datasets. Our fNIRS foundation model, MtReNIRS, jointly encodes the slow hemodynamic nature of fNIRS, chromophore identity and cortical location by mapping heterogeneous montages onto a shared anatomical reference. MtReNIRS achieves cross-subject performance on par or better than fully supervised baselines and excels in few-shot within-subject calibration, where it outperforms prior state-of-the-art fNIRS models, while remaining robust to sparse montages. Our results indicate that montage-aware self-supervision yields physiologically grounded, transferable fNIRS representations, providing a scalable foundation for label-efficient brain-computer interfaces.
Related
- Biobank-scale self-supervised genotype pretraining yields transferable representations for human genetics
- BiomECG: Learning Transferable ECG Representations from Small Cohorts
- PhysBridge: Response-supervised grid pretraining with physics-constrained local correction for transferable power-system intelligence
- Self-supervised representations reveal the genetic architecture of human cortical folding
- A self-supervised DNA foundation model with collapse-resistant multimodal fusion
BibTeX
@misc{chae2026mtrenirs,
title = {{MtReNIRS: A Self-Supervised Foundation Model for Transferable Neurovascular Representations}},
author = {Tony Chae and Kleanthis Avramidis and Dimitra Emmanouilidou and Ivan J. Tashev},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.36159},
eprint = {2610.36159},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.36159},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - MtReNIRS: A Self-Supervised Foundation Model for Transferable Neurovascular Representations
AU - Tony Chae
AU - Kleanthis Avramidis
AU - Dimitra Emmanouilidou
AU - Ivan J. Tashev
PY - 2026
DA - 2026/10/26
PB - arXiv
DO - 10.48550/arxiv.2610.36159
UR - https://arxiv.org/abs/2610.36159
AB - Functional near-infrared spectroscopy (fNIRS) offers a promising approach for monitoring the hemodynamic correlates of neural activity in naturalistic settings. However, existing modeling methods for fNIRS are largely confined to small-scale, single-session data with low sampling rates, driven by high heterogeneity in acquisition devices and optode montages. In addition, the complex spatiotemporal structure of fNIRS signals further restricts model scalability and transferability across settings. As a result, it remains unclear whether a single fNIRS encoder can learn representations that are usable across subjects, tasks, devices, and montages. Here, we show that a montage-aware masked autoencoder, pretrained on 12 public fNIRS datasets spanning 268 subjects and 567 recording hours, learns generalizable neurovascular representations that transfer across three external datasets. Our fNIRS foundation model, MtReNIRS, jointly encodes the slow hemodynamic nature of fNIRS, chromophore identity and cortical location by mapping heterogeneous montages onto a shared anatomical reference. MtReNIRS achieves cross-subject performance on par or better than fully supervised baselines and excels in few-shot within-subject calibration, where it outperforms prior state-of-the-art fNIRS models, while remaining robust to sparse montages. Our results indicate that montage-aware self-supervision yields physiologically grounded, transferable fNIRS representations, providing a scalable foundation for label-efficient brain-computer interfaces.
ER -