The field of physiological signal processing is witnessing significant advancements with the development of innovative multimodal approaches. Researchers are increasingly focusing on designing models that can effectively integrate and analyze multiple modalities, such as ECG, PPG, and accelerometry, to capture a more comprehensive understanding of physiological processes. These models are being designed to learn shared representations across modalities, enabling the discovery of complementary information and improving the overall performance of downstream tasks. Notably, self-supervised learning techniques are being explored to address the challenges of limited labeled data and missing modalities. The use of foundation models, pre-trained on large-scale datasets, is also gaining traction, allowing for efficient adaptation to various downstream tasks and accelerating scientific discovery.

Some noteworthy papers in this regard include: Leveraging Shared Prototypes for a Multimodal Pulse Motion Foundation Model, which proposes a novel SSL framework that introduces a shared prototype dictionary to anchor heterogeneous modalities in a common embedding space. Vision4PPG: Emergent PPG Analysis Capability of Vision Foundation Models for Vital Signs like Blood Pressure, which demonstrates the effectiveness of vision foundation models for PPG analysis and vital sign estimation. MCE: Towards a General Framework for Handling Missing Modalities under Imbalanced Missing Rates, which proposes a framework to tackle the challenge of missing modalities and imbalanced missing rates. SS-DPPN: A self-supervised dual-path foundation model for the generalizable cardiac audio representation, which introduces a dual-path contrastive learning-based architecture for cardiac audio representation and classification. Robust Photoplethysmography Signal Denoising via Mamba Networks, which proposes a deep learning framework for PPG denoising with an emphasis on preserving physiological information. PhysioME: A Robust Multimodal Self-Supervised Framework for Physiological Signals with Missing Modalities, which proposes a robust framework designed to ensure reliable performance under missing modality conditions. MIEO: encoding clinical data to enhance cardiovascular event prediction, which proposes the use of self-supervised auto-encoders to efficiently address the challenges of low availability of labelled data and data heterogeneity. Reconstructing 12-Lead ECG from 3-Lead ECG using Variational Autoencoder to Improve Cardiac Disease Detection of Wearable ECG Devices, which proposes a Variational Autoencoder method that reconstructs twelve-lead ECGs from three leads. DE3S: Dual-Enhanced Soft-Sparse-Shape Learning for Medical Early Time-Series Classification, which proposes a novel Dual-Enhanced Soft-Shape Learning framework to figure out shapelets precisely. On Foundation Models for Temporal Point Processes to Accelerate Scientific Discovery, which introduces a new approach: a single, powerful model that learns the underlying patterns of event data in context. One Dimensional CNN ECG Mamba for Multilabel Abnormality Classification in 12 Lead ECG, which introduces a hybrid framework named One Dimensional Convolutional Neural Network Electrocardiogram Mamba. Generalist vs Specialist Time Series Foundation Models: Investigating Potential Emergent Behaviors in Assessing Human Health Using PPG Signals, which conducts a comprehensive benchmarking study to compare the performance of generalist and specialist models.