Wearables: fatigue, depression, big growth

A team at the National University of Singapore built a soft, skin‑like hydrogel sensor that records clear heart rhythms and blood‑pressure signals while people go about normal activity, then runs those signals through an AI model that classified fatigue levels with about 92% accuracy. (nature.com) (news.nus.edu.sg) A multi‑site group led by Miriam Ina Hehlmann and colleagues combined data from university students and outpatient clinics to train an algorithm that estimates depression symptom severity from wearable signals; the model separated screen‑positive from non‑depressed participants with about 79% overall accuracy and an area under the curve of 0.82 (a standard measure of classification performance). (nature.com) On the materials side, the NUS device uses a “meta‑topological” hydrogel that pairs two filtering strategies: a programmable phononic metastructure that blocks mechanical vibrations (motion noise) and a topology‑tunable ion network that lets low‑frequency cardiovascular signals pass while damping higher‑frequency muscle interference. (nature.com) That material design is complemented by signal‑level processing: the team embedded nanoparticles in the gel to scatter vibrations and used a glycerol‑water electrolyte to favour heart‑rate frequencies, which together pushed electrocardiogram signal quality from a low baseline to 37.36 dB signal‑to‑noise and reduced blood‑pressure error to about 3 mmHg during movement. (news.nus.edu.sg) (nature.com) On the analytics side, the fatigue work used deep‑learning classifiers to remove remaining artefacts and raised peak‑detection accuracy from roughly 52% to 93% while producing the reported 92.04% fatigue‑classification score, and the depression study relied on an elastic‑net regularized logistic regression (a statistical classifier that selects the most informative features while avoiding overfitting) where the strongest predictors were minimum awake heart rate, variability in sleep duration, and maximum daily step count. (nature.com 1) (nature.com 2) The depression paper notes sample‑specific effects and restricted data availability (one clinical dataset is not shareable while analysis scripts and some project files are hosted on an open repository), and the market analysis framing these advances categorizes pharmabiotech wearables by product (smartwatches, continuous glucose meters, bands, rings, patches, injectors) and by application (drug discovery, clinical trials, medication adherence), highlighting where devices with better motion‑tolerant sensing and validated severity scores could be deployed. (nature.com) (marketsandmarkets.com) Publication dates and venues: the metahydrogel work appeared in Nature Sensors on 24 March 2026, the wearable‑based depression severity study is in Scientific Reports published 3 April 2026, and the market report was released in early April 2026, providing the proximate evidence base and commercial framing for these concurrent advances. (nature.com 1) (nature.com 2) (prnewswire.com)