The review, led by Burhan Adıgüzel and colleagues at Fatih Sultan Mehmet Vakıf University in Istanbul, systematically examines the three major sensor families that underpin modern sports wearables. Biopotential sensors, including electrocardiography (ECG), electromyography (EMG) and electroencephalography (EEG), tap the electrical activity of the heart, muscles and brain. Physical sensors, spanning temperature probes, piezoresistive and piezoelectric elements, accelerometers, gyroscopes, inertial measurement units (IMUs), pressure and optical devices, quantify movement, mechanical load, posture and thermal state. Chemical sensors, particularly sweat-based platforms, decode the biochemical signature of exertion, tracking glucose, lactate, electrolytes, pH and the stress hormone cortisol. The commercial stakes are enormous: the global wearable technology market was valued at roughly $84.2 billion in 2024 and is projected to reach $186.14 billion by 2030, a compound annual growth rate of 13.6 percent.

Biopotential sensing is the most technically mature of the three families, and recent advances are making it dramatically more wearable. Shen and colleagues demonstrated that custom-designed 3D knitted silver electrodes integrated into a smart garment could acquire ECG signals with a signal-to-noise ratio of 6.95 dB, approaching the fidelity of conventional Ag/AgCl clinical electrodes while eliminating adhesive patches. Valchinov and colleagues pushed further, building a wireless ECG system based on coin-sized dry electrodes that operate through clothing, amplify signals with a differential gain of 56 dB and transmit via ultra-low-power ANT+ technology. On the muscular side, the Athos wearable sEMG system, which embeds electrodes into compression garments, was validated against research-grade instrumentation across isokinetic knee tasks at multiple velocities and contraction intensities, showing no significant differences in amplitude metrics or the relationship between muscle activity and torque output.

Yet the review is candid about the Achilles heel of biopotential monitoring: the skin–electrode interface. Wet hydrogel electrodes dehydrate during prolonged endurance events, driving rapid impedance spikes, while dry conductive polymer or textile electrodes suffer from high baseline contact resistance. Motion-induced shear forces cause micro-displacements at the stratum corneum, generating artifacts that overlap with the frequency spectra of ECG and EMG signals. One elegant solution comes from nature: Li and colleagues developed a sweat-tolerant biopotential electrode based on conical microporous Janus gold nanowire/nitrocellulose structures, inspired by directional fluid transport in living organisms. The Janus architecture spontaneously transfers sweat from the skin-facing side into the nitrocellulose layer, maintaining conformal contact and low interfacial impedance even during prolonged, sweat-soaked measurements.

Physical sensors, meanwhile, are quietly transforming how coaches quantify the invisible mechanics of sport. Flexible piezoresistive pressure sensors built from MXene–melamine sponge composites have been integrated into smart tennis rackets, achieving a sensitivity of 5.35 kPa⁻¹ in the 1.1–22.2 kPa range and, when coupled with artificial intelligence algorithms, a stroke-classification accuracy of 98.2 percent across forehand, backhand, serve and volley. A wrist-mounted MEMS platform analyzed data from twelve elite tennis players performing 150 strokes each, with a support vector machine classifier reaching 99 percent test accuracy using an RBF kernel and Min–Max normalization. Piezoelectric sensors built from electrospun PVDF/BaTiO₃ nanocomposite fibers offer a complementary trick: they harvest mechanical energy from motion itself, powering wireless transmission of squatting, walking and running data without any external battery.

Chemical sensing is the youngest and most volatile member of the trio, but also the most tantalizing. Sweat contains a rich cocktail of biomarkers, and electrochemical platforms can now read them in real time. Vinoth and colleagues built a fully printed, pump-free microfluidic device that transports sweat to sensing electrodes within 40 seconds, simultaneously measuring lactate, sodium, potassium and pH during stationary cycling. Enzyme-free alternatives are emerging too: Pei and colleagues developed a potentiometric lactate sensor based on a multiwalled carbon nanotube/Cu-HHTP metal–organic framework composite, achieving a detection limit of 1.09 μM with good reproducibility across physiologically relevant ranges. Molecularly imprinted polymers and aptamer-functionalized DNA nanostructures are pushing selectivity further, enabling stable cortisol monitoring over 120 cycles and multivalent detection of cortisol, lactate and uric acid directly in human sweat with smartphone assistance.

The review’s most provocative argument is that multimodal integration, not individual sensor performance, is the true frontier. The authors analyze how heterogeneous signals must be temporally aligned and fused, distinguishing signal-level, feature-level and decision-level strategies. Feature-level fusion is particularly attractive: EMG descriptors such as root-mean-square amplitude and median frequency can be concatenated with sweat lactate concentration, heart rate, skin temperature and motion-derived features to characterize neuromuscular and metabolic states simultaneously. But the authors warn that simply stacking sensors on the same platform is not multimodal intelligence. Meaningful integration requires joint interpretation of complementary physiological signals to generate information that no single modality could reliably provide, and the relationship between biochemical accumulation and electrophysiological fatigue depends strongly on exercise modality, contraction type and the specific feature examined.

Artificial intelligence is the engine making this convergence actionable. A Transformer–GCN framework that synchronized physiological and behavioural signals achieved a 20.1 percent improvement in VO₂max and a 99.3 percent increase in muscular endurance after a 12-week intervention. A spatial–temporal attention-enhanced LSTM integrating heart rate, three-axis acceleration and environmental data reached 85.3 percent accuracy in exercise-intensity classification, outperforming conventional approaches by 13.2 percent. Deep reinforcement learning frameworks that optimize training load in real time have improved performance by 12.3 percent while reducing injury rates by 43 percent, with response times below 2 seconds. At the hardware level, modular platforms such as VersaSens incorporate hardware-accelerated edge-AI co-processors to sustain continuous multimodal acquisition under strict energy budgets, while TinyML architectures perform signal-quality assessment and artifact detection directly on wearable hardware rather than streaming raw data to the cloud.

The review does not shy away from the bottlenecks that separate laboratory prototypes from dependable field tools. Enzyme-based sweat sensors suffer from biofouling, pH sensitivity and thermal degradation that can shorten operational lifespans to mere hours. Epidermal platforms face a fundamental decoupling between sweat-borne analyte concentrations and systemic blood levels, compounded by fluidic latency in microfluidic channels where newly secreted sweat mixes with stale residual fluid. IMU-based systems accumulate numerical integration drift during long-duration tracking, and soft-tissue motion introduces non-linear mechanical noise that distorts impact estimates in collision sports. Validation studies reveal sobering gaps: a wrist-worn optical heart-rate monitor showed increasing errors with exercise intensity, particularly among participants with darker skin tones, while a full-marathon ECG trial achieved adequate data collection in only 65 percent of runners.

Perhaps the most important translational limitation is the performance gap between controlled laboratory validation and unrestricted athletic environments. Outdoor and competitive sports expose sensors to variable temperature, humidity, solar radiation, sweat accumulation, rain, repeated impacts and inconsistent placement, creating a distribution shift that can render algorithms trained under standardized conditions unreliable. The authors argue that future research must move beyond isolated sensitivity or classification-accuracy metrics toward standardized, longitudinal, system-level benchmarking that simultaneously evaluates signal drift, motion robustness, mechanical durability, energy consumption, calibration stability, privacy and multimodal reliability under realistic sporting conditions. Privacy deserves particular attention: physiological signals can support re-identification even after direct identifiers are removed, and federated learning, while promising, does not eliminate model inversion or gradient leakage risks. If these challenges are met, the authors conclude, multimodal wearable biosensing will transform sports monitoring from isolated gadgets into integrated platforms for performance evaluation, injury prevention and long-term health management, closing the loop between what the body does, what it feels and what it is made of.

Subject of Research: Wearable Biosensors

Article Title: Multimodal wearable biosensors in sports: integrating physical and chemical sensing for performance and health

Article References: Adıgüzel, B., Ghorbanpour, M., Sarac, B., Yücer, S., & Ciftci, F. (2026). Multimodal wearable biosensors in sports: integrating physical and chemical sensing for performance and health. Frontiers in Sensors, 7, Article 1884845. https://doi.org/10.3389/fsens.2026.1884845

Image Credits: AI Generated

DOI: 10.3389/fsens.2026.1884845

Keywords: wearable biosensors, sports technology, sweat sensors, biopotential sensors, multimodal sensing, artificial intelligence, injury prevention, flexible electronics, Multimodal, wearable, biosensors, sports