AI is being broadly used to investigate biosignals measured by wearable units akin to smartwatches. By figuring out significant patterns in well being knowledge — together with sleep, coronary heart fee, and bodily exercise — AI helps customers higher perceive and handle their well being.
The Health Forum held at Galaxy Unpacked July 2026 additionally shared Samsung’s Connected Care imaginative and prescient for the following chapter of digital well being — a future the place care strikes from reactive therapy towards preventive, customized, and related experiences, supported by trusted well being innovation and partnerships throughout the healthcare ecosystem. One expertise that’s serving to energy these new shopper experiences is the well being basis mannequin.
Researchers from the Digital Health Team at Samsung Research America (SRA) are creating new AI applied sciences to repeatedly perceive an individual’s well being state from biosignals, generate well being insights, and provide applicable well being steerage.
Samsung researchers lately launched two basis fashions based mostly on wearable knowledge: xMAE (Physiology-Aware Masked Cross-Modal Reconstruction for Biosignal Representation Learning), which learns the temporal relationship between completely different biosignals, and HiMAE (Hierarchical Masked Autoencoder), which understands well being patterns throughout completely different time scales in wearable time-series knowledge. Both research exhibit development of well being AI fashions that may higher perceive the physiological relationships and temporal construction of biosignals.
Samsung’s work on xMAE and HiMAE fashions was accepted to the International Conference on Machine Learning (ICML), and the International Conference on Learning Representations (ICLR), respectively, highlighting the importance of this analysis.
Health Foundation Models That Read the Body’s Signals: Why They Matter for Health
A well being basis mannequin is an AI mannequin that makes use of self-supervised studying to study significant options from unlabeled biosignal knowledge. After being pretrained on large-scale well being knowledge, the fashions will be utilized to carry out a variety of downstream well being duties, together with biosignal evaluation, creating new or improved biomarkers, and well being subject prediction.
xMAE AI mannequin: Continuous Cardiac Insight from Wearables
An electrocardiogram (ECG) on wearables immediately measures {the electrical} exercise of the guts and is helpful for measuring coronary heart fee and coronary heart fee variability, figuring out irregular coronary heart rhythms and dangers for circumstances akin to atrial fibrillation. ECG is very correct, however sometimes requires customers to pause and take an energetic measurement. On the opposite hand, photoplethysmography (PPG) can not directly measure cardiac operate by detecting modifications in blood circulate and will be passively and repeatedly measured by sensors in wearable units akin to smartwatches.
The two alerts originate from the identical cardiac exercise, however they seem with a sure time distinction, very like thunder is heard after lightning is seen. xMAE is a biosignal pretraining framework designed to study the temporal relationship between the 2 alerts by reconstructing masked parts of the ECG sign utilizing the extra simply and repeatedly measured PPG sign. As a outcome, cardiovascular health-related options will be analyzed extra exactly utilizing PPG with out requiring separate guide ECG measurements.
The researchers pretrained xMAE utilizing roughly 9,400 hours of ECG and PPG knowledge. The mannequin outperformed unimodal biosignal fashions and present multimodal studying strategies in 15 of 19 analysis duties, together with heart problems prediction, irregular test-result detection, and sleep-stage classification. The research additionally confirmed the potential to make use of the discovered options throughout completely different sensor units, physique places and data-gathering environments.
HiMAE AI mannequin: Reading the Body at Every Time Scale on the Device
Health knowledge measured from wearable units can reveal…
