A Comprehensive Review of Biosignal Foundation Models

Summary

This Imperial College London / Aristotle University survey, led by Konstantinos Barmpas's group, provides a comprehensive structured review of the rapidly growing field of biosignal foundation models (FMs). The review covers EEG, ECG, EMG, EOG, and PPG signals and synthesises 100+ papers on FM approaches across unimodal and multimodal settings. The paper structures its analysis around data processing, feature extraction, model architectures, pre-training paradigms (masked autoencoding, contrastive learning, self-supervised pretext tasks), and evaluation methods.