Official implementation of the paper:
SignalMC-MED: A Multimodal Benchmark for Evaluating Biosignal Foundation Models on Single-Lead ECG and PPG, 2026 [arXiv].
Fredrik K. Gustafsson, Xiao Gu, Mattia Carletti, Patitapaban Palo, David W. Eyre, David A. Clifton.
Biosignal foundation models (FMs) have shown promise for clinical prediction, yet systematic evaluation on long-duration multimodal data remains limited. We introduce SignalMC-MED, a benchmark of 22,256 emergency department visits with synchronized 10-minute single-lead ECG and PPG, evaluating FMs across 20 clinically relevant tasks. Using this benchmark, we compare representative time-series and biosignal FMs across ECG-only, PPG-only, and ECG + PPG settings. Domain-specific biosignal FMs outperform general time-series models, multimodal ECG + PPG fusion and longer signal segments consistently improve performance, larger model variants do not reliably outperform smaller ones, and hand-crafted ECG domain features remain strong complementary baselines.
If you find this work useful, please consider citing:
TODO!
Please also cite the original MC-MED dataset:
@article{kansal2025mc,
title={{MC-MED}, multimodal clinical monitoring in the emergency department},
author={Kansal, Aman and Chen, Emma and Jin, Boyang Tom and Rajpurkar, Pranav and Kim, David A},
journal={Scientific Data},
volume={12},
number={1},
pages={1094},
year={2025},
publisher={Nature Publishing Group UK London}
}
@article{PhysioNet-mc-med-1.0.1,
author = {Kansal, Aman and Chen, Emma and Jin, Tom and Rajpurkar, Pranav and Kim, David},
title = {{Multimodal Clinical Monitoring in the Emergency Department (MC-MED)}},
journal = {{PhysioNet}},
year = {2025},
month = sep,
note = {Version 1.0.1},
doi = {10.13026/wvyw-g663},
url = {https://doi.org/10.13026/wvyw-g663}
}
More detailed info will be added to this readme later...
TODO!
