MorphologyFM: A Foundation Model for Morphology-Aware Representation Learning from ECG and Pulse Oximetry Waveforms
Saiyang Feng, Yuanyun Zhang, Shi Li
Why It Matters
What makes this one worth your time
Understanding and predicting physiological conditions from ECG and SpO2 waveforms can enhance patient monitoring and care in critical settings.
MorphologyFM leverages waveform morphology for improved physiological representation learning.
Summary
The paper introduces MorphologyFM, a foundation model for learning morphology-aware representations from ECG and SpO2 waveforms using a self-supervised learning approach. The model is trained on the MIMIC critical care database and evaluated on various prediction tasks, showing improvements over existing self-supervised methods.
Key contributions
- Introduction of a morphology-aware self-supervised learning objective for waveform data.
- Demonstration of improved performance on multiple downstream tasks compared to existing methods.
- Proposal of a multimodal approach combining ECG and SpO2 waveforms for better representation learning.
Notable insights
- Morphology-guided masking and cross-modal representation learning are used to capture clinically relevant structures.
- Joint modeling of ECG and SpO2 waveforms provides more transferable representations than single modality training.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2607.09749v1 Announce Type: cross Abstract: Foundation models have recently emerged as a powerful paradigm for learning transferable representations from large scale biomedical data, yet existing approaches for physiological waveforms primarily optimize reconstruction or forecasting objectives that do not explicitly preserve clinically meaningful waveform morphology. Electrocardiograms (ECGs) and pulse oximetry (SpO2) waveforms encode rich cardiovascular and hemodynamic information through their morphological structure. In this work, we introduce MorphologyFM, a multimodal foundation model pretrained on paired ECG and SpO2 waveforms from the MIMIC critical care database using a morphology aware self supervised learning objective. MorphologyFM combines morphology guided masking, cross modal representation learning, and contrastive latent alignment to learn representations that capture clinically relevant physiological structure without requiring manual annotations. We evaluate MorphologyFM across multiple downstream prediction tasks, including arrhythmia classification, hypoxemia prediction, mortality prediction, and length of stay estimation, demonstrating consistent improvements over representative self supervised learning methods, including Masked Autoencoders (MAE), contrastive learning, Barlow Twins, and Joint Embedding Predictive Architectures (JEPA). Furthermore, we show that jointly modeling ECG and SpO2 waveforms produces more transferable representations than single modality pretraining. Our results establish waveform morphology as a powerful inductive bias for self supervised physiological representation learning and introduce MorphologyFM as a general purpose foundation model for continuous physiological monitoring.