Hollow Ti3C2Tx-MXene nanosphere electrochemical sensor coupled with LSTM time-series modeling for rapid detection of ascorbic acid in sweat

Hollow Ti 3 C 2 T x MXene nanospheres are synthesized via a sacrificial PS template to boost electroactive surface area. • The sensor achieves high AA sensing performance (0.394 μA·μM −1 , LOD 6.1 μM) covering sweat physiological concentration range. • A CNN-LSTM model is built to predict steady-state current using only 20 s transient chronoamperometric response. Real-time wearable health monitoring is frequently bottlenecked by the slow kinetic equilibrium of electrochemical sensors, which delays data acquisition and hinders high-frequency sampling. To bridge the gap between high sensitivity and rapid readout, this study presents a synergistic sensing strategy coupling hollow Ti 3 C 2 T x MXene nanospheres with a deep learning-enabled predictive algorithm for the detection of ascorbic acid (AA) in sweat. Structurally, the hollow Ti 3 C 2 T x spheres, synthesized via a sacrificial polystyrene template, provide a significantly enhanced electroactive surface area, achieving a high sensitivity of 0.394 μA·μM −1 and a limit of detection of 6.1 μM. Crucially, to resolve the trade-off between measurement accuracy and readout speed, a convolutional neural network–long short-term memory (CNN–LSTM) cascade model was constructed. The CNN–LSTM model predicted the 60 s reference current from the first 20 s transient response, reducing the effective readout time by approximately 67%. Real-sweat validation was further performed using samples collected from five volunteers after exercise, with each sample measured in triplicate. The AA concentrations calculated from the conventional 60 s current-based method ranged from 22.05 to 36.22 μM, while the CNN–LSTM-predicted concentrations showed relative errors of 3.12–11.04% compared with the conventional readout., This integration of advanced hollow nanostructures with transient-signal learning establishes a robust pathway for next-generation wearables requiring instantaneous bio-fluid analysis. Real-time wearable health monitoring is frequently bottlenecked by the slow kinetic equilibrium of electrochemical sensors, which delays data acquisition and hinders high-frequency sampling. To bridge the gap between high sensitivity and rapid readout, this study presents a synergistic sensing strategy coupling hollow Ti 3 C 2 T x MXene nanospheres with a deep learning-enabled predictive algorithm for the detection of ascorbic acid (AA) in sweat. Structurally, the hollow Ti 3 C 2 T x spheres, synthesized via a sacrificial polystyrene template, provide a significantly enhanced electroactive surface area, achieving a high sensitivity of 0.394 μA·μM −1 and a limit of detection of 6.1 μM. Crucially, to resolve the trade-off between measurement accuracy and readout speed, a convolutional neural network–long short-term memory (CNN–LSTM) cascade model was constructed. Unlike traditional methods that await steady-state current (∼60 s), this algorithm extracts deep temporal features from the initial 20 s of the transient chronoamperometric response to accurately forecast the final equilibrium signal. The model demonstrates exceptional predictive accuracy (deviation <2.8%) while reducing the detection latency by approximately 67%. Validated in human sweat samples (29.4 μM AA), this integration of advanced hollow nanostructures with transient-signal learning establishes a robust pathway for next-generation wearables requiring instantaneous bio-fluid analysis. Download: Download high-res image (117KB) Download: Download full-size image