Anti-swelling Janus hydrogels enable long-term muscle fatigue monitoring via machine learning-assisted signal analysis

Antifouling Janus hydrogel with an asymmetric architecture is fabricated and the interface is enhanced by annealing. • The Janus hydrogel exhibits a superior signal-to-noise ratio of 31 dB compared to commercial alternatives (22 dB). • The Janus hydrogel can maintain a robust 25 dB even after a 30-day immersion period. • By integrating a GBDT model, the acquired signals support autonomous muscle fatigue classification with 92.6% accuracy. Water-induced adhesion failure and swelling compromise the viability of conventional hydrogel electrodes for long-term muscle fatigue monitoring under dynamic and humid conditions. To address these limitations, we report an anti-swelling Janus hydrogel with an asymmetric architecture that simultaneously ensures robust interfacial adhesion and high signal fidelity. The design comprises an adhesive layer of acrylic acid (AA) and lauryl methacrylate (LMA) copolymer with incorporated poly(vinyl alcohol) (PVA) and a non-adhesive layer of pure PVA. Freezing initiates PVA microcrystallization to form the primary network, while subsequent heat annealing further densifies microcrystal formation at the interface. Crucially, the synergistic hydrophobic association and dense physical cross-linking restrict the swelling ratio to below 7% to prevent network dilution. The dynamic hydrogen bonding yields a robust interfacial adhesion of 67 kPa on skin to minimize motion artifacts. This dual stability directly translates into an exceptional initial signal-to-noise ratio (SNR) of 31 dB (versus 22 dB for commercial alternatives) and a sustained SNR of 25 dB even after a 30-day immersion period. When coupled with a Gradient Boosting Decision Tree algorithm, the acquired signals support autonomous muscle fatigue classification with 92.6% accuracy. This anti-swelling Janus strategy establishes a reliable material-algorithm interface for long-term wearable electromyography. Download: Download high-res image (131KB) Download: Download full-size image