The field of wearable sensors has undergone a paradigm shift over the past five years, transitioning from simple step counters and heart-rate monitors to sophisticated, multi-modal platforms capable of continuous, non-invasive biochemical and biophysical analysis. This evolution is driven by convergent advances in flexible electronics, nanomaterials, machine learning, and low-power communication protocols. Recent breakthroughs have moved the field beyond passive diagnostics toward active, closed-loop therapeutic systems, fundamentally redefining the interface between humans and digital health infrastructure.

1. Sweat and interstitial fluid: The new frontier of molecular wearables

While epidermal patches for electrocardiogram (ECG) and electromyography (EMG) have reached commercial maturity, the most striking progress lies in molecular sensing of biofluids. A landmark study by Sempionatto et al. (2021,Nature Biotechnology) demonstrated a fully integrated, multi-analyte sweat sensor that simultaneously tracks glucose, lactate, uric acid, and sodium, alongside skin temperature and heart rate. The key breakthrough was the development of a microfluidic network with passive capillary action and a "sweat-rate independent" design, which eliminates the dilution problem that previously plagued sweat analysis. By incorporating ion-selective electrodes and enzymatic amperometric sensors on a single flexible polyimide substrate, the device achieved clinically relevant accuracy (correlation coefficient >0.9 against venous blood) during prolonged exercise and heat stress.

More recently, the focus has shifted to interstitial fluid (ISF), which offers a closer proxy for blood glucose and drug concentrations. Researchers at the University of California, San Diego (Chen et al., 2023,Science Advances) introduced a microneedle patch with hollow, porous microneedles (200 µm length) that extracts ISF without pain. The patch integrates a plasmonic biosensor for real-time detection of cortisol and a reverse iontophoresis module to actively enhance ISF flux. This dual-mode approach achieved continuous monitoring for 72 hours in a human pilot study, with a drift of less than 5%. Critically, the microneedles are made of a biodegradable hydrogel composite, eliminating the risk of broken needles left in the skin.

2. Self-powered and energy-autonomous systems

A persistent bottleneck for long-term wearable deployment is power supply. Traditional lithium-ion batteries are bulky, rigid, and require frequent recharging. The recent development of triboelectric nanogenerators (TENGs) and biofuel cells has opened a path toward self-powered sensors. A notable contribution by Yu et al. (2024,Nature Electronics) describes a "sweat-powered" wearable system that uses lactate biofuel cells to generate electricity from the wearer's own sweat. The biofuel cell employs a novel 3D porous gold electrode functionalized with lactate oxidase and a tetrathiafulvalene mediator, achieving a power density of 1.2 mW/cm²—sufficient to drive a Bluetooth Low Energy (BLE) transmitter and a temperature sensor continuously during moderate exercise.

Even more ambitious is the integration of TENGs with energy storage. A recent demonstration by Luo’s group (2023,Advanced Materials) combined a soft, stretchable TENG (based on MXene-embedded silicone) with a micro-supercapacitor array. The TENG harvests mechanical energy from joint motion (e.g., wrist bending, walking) and stores it in the supercapacitor. The entire system is encapsulated in a breathable, waterproof elastomer and remains functional after 10,000 stretching cycles at 50% strain. This energy autonomy is particularly critical for implantable or near-implantable sensors, where battery replacement is impossible.

3. Artificial intelligence and on-device inference

Raw sensor data are useless without intelligent interpretation. The integration of edge computing with wearable sensors has enabled real-time anomaly detection and personalized health alerts. A milestone paper by Jiang et al. (2023,IEEE Transactions on Biomedical Engineering) presented a wrist-worn photoplethysmography (PPG) sensor embedded with a convolutional neural network (CNN) accelerator chip. The chip, fabricated in 28-nm CMOS technology, performs real-time atrial fibrillation detection with 94% sensitivity and 96% specificity while consuming only 89 µW. Crucially, all inference is performed on-device, meaning no raw physiological data leaves the body—a major privacy advantage over cloud-based processing.

Beyond classification, generative AI models are now being used to synthesize missing sensor data. For instance, a 2024 study innpj Digital Medicineused a variational autoencoder to reconstruct missing electromyography (EMG) segments from accelerometer and gyroscope data, enabling uninterrupted gesture recognition even when a sensor fails. This robustness is essential for prosthetics and human-machine interfaces.

4. Closed-loop therapeutic wearables

The most transformative development is the convergence of sensing with actuation. Wearable sensors are no longer just diagnostic tools; they are becoming autonomous therapeutic devices. The first commercial closed-loop system, the Medtronic MiniMed 780G, already automates insulin delivery based on continuous glucose monitoring (CGM). However, recent academic work has expanded this concept to other diseases.

A pioneering study by Lee et al. (2024,Nature Biomedical Engineering) described a "smart bandage" for chronic wound management. The bandage integrates: (a) a pH and uric acid sensor to detect infection, (b) a microcontroller that assesses wound state, and (c) an array of micro-needle electrodes that deliver antimicrobial peptides or growth factors on demand. In a diabetic mouse model, the bandage reduced wound closure time by 40% compared to passive dressings. The authors highlight that the closed-loop algorithm triggers drug release only when infection markers exceed a threshold, minimizing systemic side effects.

Similarly, for neurological disorders, a wearable closed-loop system for tremor suppression has been demonstrated. A 2023 study inScience Translational Medicineused a wrist-worn inertial sensor to detect tremor onset, followed by transcutaneous electrical nerve stimulation (TENS) at the median nerve. The system achieved a 62% reduction in tremor amplitude in Parkinson’s disease patients during daily activities. The latency from tremor detection to stimulation is less than 50 ms, enabled by a dedicated low-latency Bluetooth protocol.

5. Challenges and future outlook

Despite these advances, several critical challenges remain. First, biofouling and sensor drift still limit long-term accuracy. While microneedle patches have improved, continuous operation beyond 7 days remains unreliable. Second, standardization is lacking: different research groups use different calibration protocols, making cross-study comparisons difficult. Third, regulatory pathways for closed-loop devices are complex, as they combine medical devices, drug delivery, and software as a medical device (SaMD).

Looking forward, three directions appear most promising. First, multimodal fusion—combining optical, electrochemical, and mechanical sensors on a single platform—will provide a more holistic view of physiological state. Second, bioresorbable wearables that dissolve after a predefined period (e.g., 2 weeks) will eliminate the need for removal, ideal for post-surgical monitoring. Third, the integration of large language models (LLMs) with wearable data could enable conversational health coaching, where the sensor data informs an AI that provides natural-language feedback and interventions.

In summary, wearable sensors have evolved from passive monitors to active, intelligent, and therapeutic systems. The convergence of flexible materials, energy harvesting, on-device AI, and closed-loop actuation is not merely incremental—it represents a new era of personalized, continuous, and proactive medicine. The next decade will likely see these systems become as ubiquitous as smartphones, transforming chronic disease management, rehabilitation, and preventive health at a population scale.

References (selected):