The research, led by Peng Yang, Haifeng Zhang, Chenglong Feng and colleagues at Shanghai University of Engineering Science and Shanghai Yangzhi Rehabilitation Hospital affiliated with Tongji University, tackled one of the thorniest problems in wearable health monitoring: a model trained on one person’s muscle signals usually fails when applied to another. This cross-participant generalization problem has limited the clinical usefulness of electromyography-based fatigue detection for years, because every individual’s signal amplitude, electrode placement, and muscle anatomy differ enough to fool conventional machine learning classifiers.

To build their dataset, the team recorded twelve-channel surface electromyography from 42 participants, 28 healthy adults and 14 people with chronic low back pain, during rehabilitation-relevant tasks. Surface electrodes capture the summed electrical activity of underlying muscles through the skin, and as muscles fatigue, the frequency content of those signals shifts measurably downward. The researchers labeled fatigue using a dual-criterion approach based on the median frequency of the signal, a standard marker of muscle fatigue onset, and then evaluated their models with leave-one-subject-out cross-validation, meaning the model was repeatedly tested on participants it had never seen during training.

The core innovation lies in what the network was forced to learn. Rather than letting a deep neural network freely mine patterns from the data, the team embedded three biomechanical soft constraints drawn from established physiology: the Hill force-velocity relationship describing how muscle force output changes with contraction speed, a fatigue-dynamics steady-state model capturing how fatigue accumulates and recovers over time, and an EMG-force residual linking electrical activity to mechanical output. A fourth regularization term enforced smoothness across the ordering of the twelve recording channels. These constraints act like guardrails, penalizing predictions that fit the data but violate the physics of how muscles actually behave.

This physics-informed neural network, or PINN, approach represents a growing trend in biomedical machine learning. Pure data-driven models can achieve impressive benchmark scores yet behave implausibly when pushed outside their training distribution, a serious concern in clinical settings. By encoding known physical laws directly into the loss function, the Shanghai team’s model had to produce outputs consistent with muscle mechanics, which the authors argue improves both generalization to unseen participants and the interpretability of the learned representations.

The performance numbers are notable. The PINN achieved an area under the precision-recall curve of 0.923 and an F1 score of 0.919, with perfect recall, meaning it never missed a true fatigue event in the tested cohort. Its PR-AUC exceeded the strongest baseline, a support vector machine, by 0.029, a difference that was statistically significant with a 95 percent bias-corrected and accelerated confidence interval of 0.015 to 0.044 and a p-value of 0.004. Convolutional neural network and transformer baselines were also outperformed, suggesting that the biomechanical priors provided information those architectures could not extract from the signals alone.

Just as important as raw accuracy is calibration, the alignment between a model’s stated confidence and its actual correctness. A classifier that says it is 90 percent confident should be right about 90 percent of the time, especially when its output will guide clinical decisions. The researchers applied temperature scaling, a post-hoc calibration technique, which reduced the expected calibration error from 0.0508 to 0.0216 and improved the Brier score from 0.1329 to 0.1305. They then converted calibrated channel probabilities into discrete fatigue-risk states using a dual-threshold hysteresis scheme with consecutive-window confirmation, a design that prevents the system from flickering erratically between fatigue states when signals hover near a decision boundary.

The constraints paid off in physical consistency as well. Compared with an otherwise identical model lacking the prior terms, the exceedance rates for the three biomechanical residuals dropped by 24.1, 23.5, and 25.7 percentage points respectively, while the channel-order residual fell by 20.0 percentage points. In plain terms, the informed model’s predictions violated the encoded laws of muscle mechanics far less often, which the authors interpret as evidence that the priors genuinely shaped the network’s internal representations rather than merely adding a regularizing penalty.

Perhaps the most clinically forward-looking element is the sequential trigger-state generation. In a chronological replay evaluation, each fatigue-risk state was updated using only the current and preceding time windows, mimicking real-time deployment on a wearable device or rehabilitation platform. This streaming design, combined with the hysteresis logic, demonstrates a feasible pipeline for continuous, closed-loop fatigue monitoring in which a system could alert a therapist or adjust exercise intensity the moment a patient’s muscles cross a calibrated risk threshold.

The authors are careful to frame the scope of their claims. The gains were demonstrated within the studied cohort, task set, and comparator models, and the study was conducted at a single rehabilitation hospital under ethics approval from the Shanghai Yangzhi Rehabilitation Hospital Medical Ethics Committee, with all participants providing written informed consent. They emphasize that prospective clinician-in-the-loop studies are still required to establish whether the technology delivers measurable clinical benefit in closed-loop rehabilitation practice. The work was supported by the National Natural Science Foundation of China and a national clinical key specialty construction project. Even with those caveats, the study offers a compelling template for the next generation of rehabilitation wearables: neural networks that not only fit the data but also obey the physics of the body they are watching, and that know when to admit uncertainty. For the millions navigating chronic low back pain, a machine that reliably recognizes the moment muscles begin to fail could turn rehabilitation from an exercise in guesswork into a precisely dosed therapy.

Subject of Research: Physics-informed neural networks for surface electromyography-based muscle fatigue monitoring in chronic low back pain rehabilitation

Article Title: Biomechanics-informed PINN with calibrated trigger-state generation for sEMG fatigue monitoring in chronic low back pain rehabilitation

Article References: Yang, P., Wang, Z., Niu, W., Wang, Y., Zhang, H., & Feng, C. (2026). Biomechanics-informed PINN with calibrated trigger-state generation for sEMG fatigue monitoring in chronic low back pain rehabilitation. BioMedical Engineering OnLine. https://doi.org/10.1186/s12938-026-01637-z

Image Credits: AI Generated

DOI: 10.1186/s12938-026-01637-z

Keywords: surface electromyography, muscle fatigue, physics-informed neural network, chronic low back pain, rehabilitation, biomechanics, probability calibration, machine learning, Hill force-velocity model, wearable health monitoring, trigger-state generation, cross-participant generalization

The research, published in Physiological Reports, was motivated by a clear problem in aging biology. As people get older, their neuromuscular system changes in ways that erode both strength and endurance. Muscle mass declines, type II fast-twitch fibers are lost in disproportionate numbers, calcium handling within muscle fibers becomes less efficient, motor unit firing rates slow, and the microvascular networks that deliver oxygen and clear metabolic waste deteriorate. Together, these changes mean older adults fatigue more quickly during repetitive tasks, which translates into real-world consequences: shorter walking endurance, poorer balance, reduced physical activity, and an elevated risk of falls and loss of independence.

Because so many of these age-related impairments are temperature-dependent, the researchers reasoned that selectively warming the quadriceps might offset some of the damage. Heating muscle above resting temperatures accelerates cross-bridge cycling, the molecular process by which myosin heads pull on actin filaments to generate force. It also speeds the release and reuptake of calcium by the sarcoplasmic reticulum, improves nerve conduction velocity, and boosts local blood flow and perfusion. Improved perfusion matters because fatigue during repeated maximal contractions is driven substantially by the accumulation of metabolic byproducts such as inorganic phosphate and hydrogen ions, which interfere with force production. If heating enhances oxygen delivery and metabolite clearance, perhaps it could delay the onset of fatigue in aged muscle, which is precisely where perfusion is most compromised.

To test this, the team recruited fifteen healthy older adults, eight of them women, with an average age of 68, all free of known illness, neuromuscular disorders, or heat intolerance. Each participant visited the laboratory once, having abstained from heavy exercise, caffeine, supplements, and alcohol for 24 hours. One thigh was wrapped in a custom garment that circulated water at 50 degrees Celsius, combined with a survival blanket, for a full 90 minutes. The contralateral limb served as a thermoneutral control. Crucially, the intervention was localized: by heating only the thigh, the researchers avoided raising core body temperature, which is known to impair performance by reducing central drive from the nervous system, a confound that may have obscured benefits in earlier whole-body heating studies.

The heating protocol worked exactly as intended. Skin temperature over the vastus lateralis rose from about 30.5 to 39.3 degrees Celsius in the heated limb, while intramuscular measurements in a small subset of volunteers showed muscle temperature climbing roughly five degrees, from about 32.2 to 37.2 degrees Celsius. Tympanic temperature and heart rate remained unchanged, confirming that the heating stayed local, and mean arterial pressure actually fell slightly. Surface electromyography of the vastus lateralis, recorded during the exercise tests, showed no significant changes in neural drive, indicating that the intervention neither boosted nor blunted the nervous system’s output to the working muscle.

The muscle function test was a demanding one. Participants completed 30 maximal knee extensions on an isokinetic dynamometer at 180 degrees per second, kicking as hard and as fast as possible through a 75-to-175-degree range of motion, both before and after the heating period. Peak torque, averaged torque across the first, middle, and final thirds of the 30 repetitions, and total work done were all analyzed. As hypothesized, the task was genuinely fatiguing: average torque fell from about 68 newton-meters in the first ten repetitions to 57 in the middle third and 51 in the final third.

Heating delivered a clear win on the single strongest contraction. Peak torque in the heated limb rose by about 8 percent after 90 minutes of warming, an increase of roughly 10 newton-meters compared with the control limb, a statistically significant difference. The control limb showed no meaningful gain. This result aligns with a growing body of evidence that passive limb heating acutely enhances peak force production, maximal power output, and even six-minute walk distance in some populations, likely through faster cross-bridge kinetics and improved calcium handling at the level of the muscle fiber.

But the hoped-for protection against fatigue never materialized. Total work done across the 30 repetitions increased by about 7 percent in the heated limb, but this change was not statistically significant, especially against a 4 percent decline in the control limb that fell within the study’s variability. There was no interaction between heating and the decline in torque across the repetition blocks, meaning the heated muscle fatigued at essentially the same rate as the unheated one. The finding echoes a 2008 study in young adults that used whole-body heating and likewise found no improvement in dynamic fatigue resistance despite higher peak forces, suggesting that whether heat is applied locally or across the whole body, it simply does not alter how quickly force output decays during repeated maximal efforts.

Why does warmth strengthen the muscle yet fail to preserve it under fatigue? The authors point to the mechanics of maximal contraction. During near-maximal efforts, the intense pressure generated inside the working muscle can effectively occlude its own blood supply, meaning any improvement in perfusion from heating cannot be exploited precisely when it would matter most. The benefits of enhanced blood flow may also unfold over longer timescales than a burst of 30 repetitions lasting less than a minute. Meanwhile, if heating does facilitate calcium cycling within the fiber, that advantage appears either too small or too easily overwhelmed by metabolic disruption to change the fatigue trajectory. The electromyography data hinted at a modest decline in neural drive as the task progressed, consistent with some contribution from central fatigue, but heating did not modulate that pattern either.

The study had limitations worth noting. A parallel exercise protocol ran between the two fatigue tests, altering muscle temperature and metabolic state, particularly in the control limb, and the control test was always performed second, raising the possibility of residual fatigue. Intramuscular temperature was measured in only three participants, and surface EMG provided only a coarse view of activation. Yet the broader message is robust and practically important: the ergogenic effects of passive heating appear to be task-specific. Prior work from the same group and others shows benefits for early force production, peak torque, and, in clinical populations with conditions like peripheral artery disease, submaximal endurance tasks such as walking. What localized heating does not do, at least in healthy older adults, is rescue performance during short, all-out, repetitive maximal efforts. For aging populations, that suggests passive heating may be best positioned as a tool to prime muscle before everyday submaximal activities, rather than a shield against high-intensity fatigue, and future research should explore whether clinical groups with impaired perfusion stand to gain more, particularly for sustained tasks where improved blood flow can actually be recruited.

Subject of Research: Effects of localized passive thigh heating on neuromuscular fatigue resistance and peak torque in older adults

Article Title: Passive thigh heating increases peak torque but does not attenuate declines in force production during repeated isokinetic knee extensor exercise in older adults

Article References: Denny, D., Low, D. C., & Gibson, O. R. (2026). Passive thigh heating increases peak torque but does not attenuate declines in force production during repeated isokinetic knee extensor exercise in older adults. Physiological Reports, 14(18), Article e71105. https://doi.org/10.14814/phy2.71105

Image Credits: AI Generated

DOI: 10.14814/phy2.71105

Keywords: passive heating, muscle fatigue, older adults, peak torque, isokinetic exercise, knee extensors, sarcoplasmic reticulum, perfusion, neuromuscular function, electromyography, aging, ergogenic aid