Postoperative pain after musculoskeletal surgery impedes early rehabilitation and reduces quality of life. Electroacupuncture is used as an adjunct for analgesia, but the optimal stimulation frequency and short-term response trajectory have not been established. The aim of this study is to compare frequency-dependent relative analgesic responses to bilateral ST36 to SP9 electroacupuncture at 2 Hz and 100 Hz and generate effect size and variability estimates to inform future definitive trials. This single-center, frequency-masked, active-comparator, 2 × 2 prospective randomized crossover trial will enroll 30 adults (aged 19-80 years) experiencing pain (numerical rating scale [NRS] score≥4) after musculoskeletal surgery. Participants will undergo two 15-minute electroacupuncture sessions at bilateral ST36 to SP9 (2 Hz and 100 Hz) separated by a washout period of at least 7 days. Participants will not be informed of the assigned stimulation frequency or treatment sequence. Participant recruitment began in September 2025, and data collection is expected to be completed by August 2026. Data analysis will begin after database finalization in late 2026, with primary findings submitted for publication in April 2027. The primary outcomes are baseline-adjusted changes in pain NRS scores at 15 minutes and 2 hours after each intervention. Secondary outcomes include pain NRS scores immediately after and at 1, 4, and 24 hours after electroacupuncture; pain location and quality; photoplethysmography-derived pulse rate variability, used as a surrogate for heart rate variability; radial pulse tonometry parameters; and wrist-worn activity metrics. This protocol describes an active-comparator crossover trial designed to estimate relative immediate and short-term analgesic responses to 2-Hz vs 100-Hz electroacupuncture and associated physiological measures in postoperative musculoskeletal pain. Because the trial does not include a sham, usual care, or no-additional-electroacupuncture control arm, it will not determine the absolute efficacy of electroacupuncture or separate specific effects from nonspecific contextual or placebo effects. Results from this trial will help optimize stimulation parameters and outcome assessments for larger, definitive efficacy trials. Clinical Research Information Service KCT0011009; https://tinyurl.com/54phyep7. DERR1-10.2196/91706.

BackgroundA key metric utilized by coaches and athletes to track athlete performance is heart rate variability (HRV). HRV calculated via electrocardiogram (ECG) has been shown to track autonomic function. However, many wearable devices utilize photoplethysmography (PPG) to calculate pulse rate variability (PRV). Our study investigated the agreement between PRV and HRV in a beat-to-beat analysis in a sample of American football players. Data from 103 male, Division I collegiate American football athletes, collected over three seasons, were analyzed. Heart rate (HR), pulse rate (PR), and two time-domain indices for PRV/HRV were measured (rMSSD and SDNN). Agreement between PRV and HRV was assessed using Bland-Altman analysis (bias, limits of agreement, and confidence intervals) with supporting error metrics (MAE, RMSE, Pearson r, and Lin’s concordance correlation coefficient). To evaluate whether PRV detected autonomic deviations of the same magnitude to HRV on the same day, a sliding-window z-score threshold-crossing analysis (0.5–2.0 SD) quantified PRV detection rates and PRV delay.ResultsHR and PR were similar at (59.4 (10.3) bpm vs. 59.7 (10.3) bpm). In contrast, PPG-PRV values were lower than ECG-HRV for both rMSSD and SDNN (80.9 (23.1) ms vs. 103.9 (22.0) ms, 141.3 (41.7) ms vs. 167.9 (40.0) ms). Bland–Altman analysis showed negative bias for rMSSD and SDNN across PPG wavelengths, race, and obesity, while HR/PR agreement was high with small bias (0.24–0.44 bpm). In the deviation analysis, PRV detected fewer autonomic deterioration events on the same day as HRV, capturing 16.7% to 56.7% of HRV-identified rMSSD events and 16.8%-52.0% of SDNN events across thresholds, and exhibited an average delay of 1.8–5.5 days in detecting changes already identified by ECG-HRV.ConclusionsOur findings refute PPG-PRV as an equivalent surrogate for ECG-HRV for tracking autonomic function over time. Specifically, the delay in response when using PPG-PRV to track athlete autonomic function will result in missed opportunities for coaches to prevent autonomic deterioration. Finally, the PRV calculated with PPG should not be called HRV as it confuses scientists and consumers.

Cardiac autonomic dysfunction is a prevalent yet underdiagnosed non-motor manifestation in Parkinson disease (PD), and difficult to monitor in daily life. Wrist-worn photoplethysmography (PPG) allows for continuous pulse rate monitoring, with potential to measure cardiac autonomic dysfunction in PD. However, motion artifacts pose significant challenges. We propose a two-step approach to account for PPG motion artifacts, and explore the association between pulse rate characteristics and cardiac autonomic dysfunction in PD. In 444 people with early PD, we used continuous wrist PPG and accelerometer data, collected during two weeks using a one research-grade device model (Verily Study Watch), with an average wear time of 22h/day. Step 1 filters low-quality segments based on PPG morphology using a logistic regression classifier. Step 2 removes remaining segments with periodic motion artifacts (e.g. introduced by tremor). Weekly aggregates were calculated for resting (daytime, nighttime) and maximum (daytime) pulse rates. We studied the effect of filtering on the pulse rate estimates and assessed the relationship between pulse rate aggregates and both subjective and objective autonomic dysfunction measures. Median proportion of high-quality PPG data was 29.2% [IQR: 24.0% to 35.9%] during the daytime, and 86.1% [IQR: 79.3% to 90.6%] during the night. Proportions of high-quality PPG data were similar across different rest tremor severity and dyskinesia scores. However, filtering out periodic motion artifacts (step 2) reduced overestimation of the maximum HR in persons with mild or severe rest tremor. More symptoms of autonomic dysfunction were associated with a lower maximum pulse rate(β:-0.17, 95% CI [-0.33, -0.00]), while resting pulse rate showed no association. No significant differences in pulse rate metrics were found between individuals with or without orthostatic hypotension. PPG-based heart rate estimation in PD improves when periodic motion artifacts are accommodated, representing an important step toward developing digital biomarkers of cardiac autonomic dysfunction. Maximum pulse rate in daily life only demonstrated a weak association with autonomic dysfunction, highlighting the need for more specific markers such as pulse rate variability.

Systemic hypertension (HTN) is a major cardiovascular comorbidity in patients with obstructive sleep apnea (OSA), yet these conditions are often diagnosed and managed independently, limiting integrated cardiovascular risk assessment. There is a significant gap in scalable automated frameworks for the screening of hypertension using routinely collected polysomnography (PSG) signals. To address this, we introduce FusionNet, a novel deep learning architecture to detect established HTN in patients with OSA. The proposed framework leverages a Multiple Instance Learning (MIL) approach to capture long-term temporal dynamics from overnight PSG recordings. It integrates electrocardiogram (ECG) and photoplethysmogram (PPG) signals through a gated fusion mechanism for robust multimodal analysis, further enriched by frequency-domain heart and pulse rate variability (HRV/PRV) and clinical context features, including sleep stages and apnea events. The model was evaluated using Leave-One-Patient-Out (LOPO) cross-validation on a clinically annotated dataset of 33 patients to ensure subject-independent generalizability. FusionNet achieved an AUROC of 87.34%, accuracy of 79.03%, specificity of 81.25%, and an MCC of 0.5801, indicating a balanced and clinically conservative detection profile through adaptive integration of complementary cardiac and vascular information. This work establishes a proof-of-concept for a dual diagnostic paradigm, demonstrating the feasibility of deriving hypertension-related cardiovascular signatures from standard sleep studies. These findings present an initial step towards automated cardiovascular risk assessment and require validation in larger independent cohorts prior to clinical deployment.

With the continued rise in outpatient surgical procedures, modern medicine requires more advanced tools for pain and anxiety monitoring and management. The current standard of care requires patient responses on visual analog scales, which may be subjective and are difficult to assess when a subject is unresponsive. Electrodermal activity (EDA) and pulse rate variability (PRV), two non-invasive, wearable, and objective measurements of sympathetic nervous system activity, can help provide insight into a patient’s psychological or emotional state without user input, allowing for continued monitoring even when a patient is unable to respond. However, methods based on these measurements have largely been relegated to longer duration (>60 s) or post hoc analysis, which does not suit the needs of medical care environments. Here we propose new methods for handling ultra-short (<10 s) signals to allow rapid evaluation of pain and anxiety state. We show how machine learning models trained on these signals can obtain high degrees of classification performance (AUC > 0.88) between no pain or anxiety and medium or higher pain and anxiety on signals obtained during two different forms of painful stimulation. We also show how these signals can measure the degree of stimulation irrespective of perceived pain from the patient. Further development of these algorithms will allow for greater monitoring and control of patient comfort in a clinical setting.

Background/Objectives: The aim of this pilot study was to evaluate the feasibility of developing individualized machine learning models using nocturnal wearable-derived autonomic nervous system (ANS) and sleep metrics to predict next-day headache risk in patients with migraine. We also examined the associations between nocturnal ANS and sleep measures and patient-reported outcome measures (PROMs) related to nociplastic pain, migraine burden, and non-restorative sleep (NRS). Methods: Adults with migraine wore the wrist-worn Empatica EmbracePlus® wearable during sleep and completed daily headache diaries for approximately 4 weeks (N = 10). Participants also completed daily headache diaries and PROMs assessing nociplastic pain, migraine burden, and non-restorative sleep. Personalized machine learning (ML) models were developed to predict next-day headache using nocturnal ANS activity (e.g., pulse rate variability (PRV), electrodermal activity (EDA), respiratory rate (RR)) and sleep metrics (e.g., interruptions, duration, awakenings). Model performance was evaluated using area under the receiver operating characteristic and precision–recall curves (AUROC, AUPRC), sensitivity, specificity, accuracy, and precision. Spearman correlations assessed the relationship between wearable-derived metrics and patient-reported outcome measurements of sleep quality (PROMIS-Fatigue, PROMIS-Sleep Disturbance) and a surrogate marker of nociplastic pain (Fibromyalgia (FM) Score). Results: 9 out of 10 participants wore the EmbracePlus device for at least the target duration of four weeks. For the next-day headache prediction, model performance varied between individuals; area under the ROC curve (AUROC) ranged from 28.2% to 81.2%. Nocturnal measures of EDA were strongly correlated with the FM score (Spearman’s rho = 0.72–0.75, p < 0.05). Conclusions: Phasic EDA may warrant further investigation as a potential physiological indicator related to nociplastic pain mechanisms and next-day headache. However, these findings are preliminary, and larger multicenter trials are needed to confirm results of this pilot study.

Currently, most studies on mental stress evaluation mainly focus on classification tasks, while research on accurately estimating continuous stress levels using deep learning for early identification remains limited. This study proposes an end-to-end continuous stress assessment framework based on a deep hybrid learning architecture. The framework employs efficient channel attention convolution to extract local pattern features from the signals, utilizes a bidirectional long short-term memory (BiLSTM) network to model contextual dependencies, and incorporates emotional cross-attention to assign importance weights to different emotional states. In addition, an adaptive ridge stacking ensemble learning method is proposed. To enhance feature representation, pulse rate variability (PRV) and discrete pulse signals (dPS) extracted from Photoplethysmography (PPG) signals are encoded into markov transition field (MTF) and recurrence plot (RP) images, respectively. The results show that, for PRV, MTF- and RP-based representations reduce the detection error by 6.93% and 2.57% compared with the time-domain baseline. For dPS, the error reductions reach 6.97% and 15.05%. Furthermore, the proposed fusion strategy of PRV-MTF and dPS-RP achieves the best performance (MAE = 3.29, RMSE = 4.05). Compared with the previous state-of-the-art method based on time-domain fusion of PRV and dPS signals (MAE = 4.38, RMSE = 5.19), the proposed approach yields substantial reductions of 24.88% in MAE and 21.96% in RMSE, reaching the current state-of-the-art performance. These results demonstrate that transforming time-domain signals into structured encoding images enables more effective capture of deep patterns associated with psychological states, thereby significantly improving the accuracy of mental health detection.

HighlightsWhat are the main findings?Evaluating the performance of individual subject models and a general model.Selecting appropriate model orders for different individual subject models and the general model.Evaluating whether different control breathing rates (CBRs) generated the respiratory sinus arrhythmia (RSA) energies coupled in the PPI signals.Evaluating whether the proposed individual subject ARMA model and the general ARMA model significantly attenuate RSA energy in the frequency-domain parameters of PRV analysis.What are the implications of the main findings?The performance of individual subject models does not significantly differ from that of the general model for removing the RSA energy coupled in the raw PPI signals.The general model could be embedded in an edge computing system in the future.Background: Pulse rate variability (PRV), a critical biomarker of autonomic nervous system (ANS) function, is typically evaluated using the pulse-to-pulse interval (PPI) signal extracted from a photoplethysmogram (PPG). Although PPGs have been widely used in wearable devices, the PPI signal is easily affected by motion artifacts or respiratory sinus arrhythmias (RSAs). These disturbances affect the accuracy of PRV for evaluating ANS function. The aim of this study was to remove the respiratory signals from raw PPI signals with an autoregressive moving average (ARMA) model. Methods: An R-wave to R-wave interval (RRI) sequence was extracted from the electrocardiogram (ECG). A self-made measurement system was used to record PPG, ECG, and respiratory signals. Nineteen healthy adults were recruited and requested to breathe with a spontaneous breathing rate (SBR) and control breathing rates (CBRs) (6, 18, and 30 breathing rate per minute, BRPM). Their ECG, PPG, and breathing signals were recorded for 6 min under different CBRs. The measurement was performed twice, i.e., eight measurements were performed. The raw RRI(t) and PPI(t) signals of 4 Hz were segmented into samples of one minute and shifted by 30 s. Thus, a subject had 80 samples, and there were 10 samples for each BRPM. RSA-free RRI signals were generated by a spectral method to filter RSA from raw RRI(t) to produce the target RRI(t). We proposed the individual subject ARMA models trained by samples with the maximum mean absolute errors between the target RRI(t) and raw PPI(t) (MAERAWs) of each subject, and the general model trained by samples with all maximum MAERAWs of all 19 subjects. Results: The mean absolute errors between the target RRI(t) and predicted by the individual subject ARMA models (MAESubject-Models) and general ARMA model (MAEGeneral-Model) were used to evaluate the performance of the two models. The results for the MAESubject-Models and MAEGeneral-Model were 132.5 ± 59.1 ms and 137.8 ± 67.8 ms, respectively, with no significant difference. MAESubject-Models and MAEGeneral-Model were compared with MAERAWs, whose attenuations (ATTs) were 28.5 ± 13.1% and 27.8 ± 12.6%, respectively. Conclusions: The two proposed models are capable of removing the RSA energy coupled in the raw PPI signals.

Objective.Pulse rate variability (PRV), derived from photoplethysmography, offers a scalable approach to studying cardiac autonomic regulation. However, the reliability of PRV measures across human lifespan has not been systematically evaluated, limiting their application in cardiovascular research.Approach.Pulse wave data of 914 healthy young subjects from the Human Connectome Project (HCP) young adults were analyzed. PRV was derived from four 15 min resting-state functional MRI sessions using NeuroKit2. Reliability of 86 PRV metrics across time, frequency, and nonlinear domains was assessed using intraclass correlation coefficients (ICCs). Findings were validated in independent HCP developmental (HCP-D;n= 536; ages 5-21) and HCP aging (HCP-A;n= 652; ages 36-100+) cohorts to evaluate lifespan generalizability. Sex-specific and race-specific analyses were also performed.Main results.We analyzed 914 healthy young adults (mean age 28.7 ± 3.7 years;n= 489, 53.5% female;n= 425, 46.5% male), of whom 75.6% self-identified as White, 13.1% as African American, and 11.3% as other. Of 86 PRV metrics, 62 (72.1%) showed excellent reliability (ICC ⩾ 0.75) across all four sessions spanning two days, with higher stability within-day than across days. Time-domain measures (e.g. Prc80NN, pNN20, MeanNN, MedianNN) and nonlinear measures (e.g. ShanEn, IALS, and PSS) were highly reliable with ICC ⩾ 0.90. In the frequency domain, HFn and LFn performed best (ICC ⩾ 0.85). The reliability patterns were highly consistent across developing and aging cohorts and were broadly similar across sex and race subgroups among highly reliable PRVs.Significance.A substantial proportion of PRV metrics showed excellent reliability across the lifespan, with broadly similar patterns across sex and racial groups, supporting their use in research settings.

HighlightsWhat are the main findings?Two simulated driving paradigms designed to induce active and passive fatigue showed distinct stage-dependent prefrontal multimodal physiological signatures.Active fatigue showed more continuous EEG changes and clearer averaged hemodynamic modulation, whereas passive fatigue showed weaker averaged hemodynamic effects and clearer pulse-variability accumulation.What are the implications of the main findings?A three-level fatigue framework helps distinguish early, late, and cumulative physiological changes that are not captured by binary fatigue classification.Multimodal prefrontal monitoring combining EEG, fNIRS, and pulse-related measures may support graded fatigue assessment and more refined fatigue monitoring in driving.Background/Objectives: Mental fatigue during driving can arise under different task conditions and typically progresses from mild to severe states. Active fatigue is usually linked to cognitively demanding driving, whereas passive fatigue is associated with prolonged monotonous driving. However, studies on multilevel mental fatigue remain scarce, and direct comparisons of prefrontal multimodal physiological responses to active and passive fatigue are still limited. The objective of this study is to characterize and compare the prefrontal multimodal physiological signatures across three fatigue levels under two simulated driving paradigms designed to induce active and passive fatigue. Methods: Eleven healthy participants completed two simulated driving tasks designed to induce active and passive fatigue. Physiological data were recorded using a self-developed prefrontal EEG-fNIRS system, and pulse-related signals were derived from the hemodynamic measurements. Based on subjective and objective indicators, fatigue was classified into non-fatigue (NonF), moderate fatigue (ModF), and severe fatigue (SevF). Results: In the active-fatigue-inducing paradigm, significant changes in prefrontal EEG and hemodynamic already emerged from NonF to ModF; for example, the EEG β/(θ + α) power ratio increased from 0.973 to 1.157 (p < 0.001) and the normalized mean deoxyhemoglobin feature increased from −0.06 to 0.09 (p < 0.001). In the passive-fatigue-inducing paradigm, EEG changes became prominent mainly from ModF to SevF, with β/(θ + α) power ratio decreasing from 0.806 to 0.761 (p < 0.05). Pulse rate variability showed increasing trends in both paradigms. Conclusions: These findings suggest that the two simulated driving paradigms were associated with distinct prefrontal electrophysiological, hemodynamic, and autonomic evolution patterns across three fatigue levels, supporting graded fatigue assessment and multimodal fatigue monitoring in driving.

Long-term variability in physiological measures and risk of frailty: evidence from two cohort studies.

This study systematically evaluates the consistency and applicability limits of photoplethysmography (PPG)-derived pulse rate variability (PRV) versus electrocardiogram (ECG)-derived heart rate variability (HRV) in real-world settings. It integrates three methodological dimensions: 24-hour multi-context monitoring, dual-level consistency analysis (inter- and intra-individual), and controlled motion intensity via a 27-level acceleration gradient. Data from 14 healthy participants were collected using synchronized wrist-worn PPG, portable ECG, and triaxial accelerometry. Standardized preprocessing and motion artifact suppression based on acceleration thresholds enabled the extraction of time-domain, frequency-domain, and nonlinear HRV and PRV metrics. Consistency was assessed using Pearson correlation and root mean square error, with false discovery rate-corrected significance testing. Results show strong PPG-ECG agreement during sleep (r &gt; 0.9 for HR, MeanNN, Prc80NN) but marked degradation under high motion. Notably, Prc20NN demonstrated exceptional robustness across contexts, retaining significant correlation even during active phases. Stringent motion filtering substantially improved correlations. These findings delineate metric-specific validity boundaries for wearable PRV, distinguishing motion-induced errors from inherent physiological discrepancies, and offer evidence-based recommendations for deploying PRV in context-appropriate applications such as sleep monitoring, passive health tracking, and longitudinal stress assessment.

Non-invasive cardiovascular risk stratification in type 2 diabetes: a pulse wave and pulse rate variability analysis with machine learning

Although pulse rate variability (PRV) is considered a potential surrogate marker for heart rate variability in the assessment of autonomic function, it is not clear whether PRV-derived parameters predict mortality risk in the general population. Between 2009 and 2018, a total of 5,943 Japanese individuals, aged 30-79 years, were recruited for a prospective study and followed until the end of 2022. The pulse wave was examined over a 5-min period using a fingertip photoplethysmography sensor to determine the resting heart rate (RHR) and the time and frequency domains of PRV. A Cox proportional hazard model was used to calculate the hazard ratio (HR) and 95% confidence interval (CI) of PRV-derived parameters for all-cause mortality using the penalized cubic splines method. During 12.4 years of follow up, 437 deaths were recorded. The HR for mortality, adjusted for sex, age, and community, for the lowest quartiles of the standard deviation of the normal-to-normal intervals (SDNN) increased 1.51 times (95% CI 1.15-1.98) vs. the third quartile. Although mortality risk was attenuated after adjustment for several confounders and RHR, PRV-derived parameters of autonomic function showed significant non-linearity of association with mortality risk in the spline analysis. Low values of PRV-derived autonomic parameters were associated with an increased risk of all-cause mortality in the general Japanese population.

Camera-based contactless monitoring of vital signs, also known as imaging photoplethysmography (iPPG), has seen applications in driver-monitoring, perfusion assessment, affective computing, and more. iPPG involves sensing the underlying cardiac pulse from video of the skin and estimating vital signs such as the pulse rate or a full pulse waveform. Some previous iPPG methods impose model-based sparse priors on the pulse signals and use iterative optimization for pulse wave recovery, while others use end-to-end black-box deep learning methods. In contrast, we introduce methods that combine signal processing and deep learning methods in an inverse problem framework. Our methods estimate the underlying pulse signal, pulse rate, and pulse rate variability from facial video by learning deep-network-based denoising operators that leverage deep algorithm unfolding and deep equilibrium models. Experiments show that our methods can denoise an acquired signal from the face and infer the correct underlying pulse rate and pulse rate variability, achieving pulse rate estimation performance consistent with the state-of-the-art on well-known benchmarks, all with less than one-fifth the number of learnable parameters as the closest competing method.

Collar monitoring devices are used in animals for the minimally invasive collection of physiological data, using software and algorithms to provide general health trends. There is potential to utilise the raw data collected from these devices to improve animal monitoring strategies and intervention points in animal disease studies. We aimed to develop an algorithm for the early detection of highly pathogenic African swine fever disease in research pigs (Sus scrofa), using data collected via modified PetPaceTM health monitoring collars. Pigs from two other studies (n = 6 per study, total n = 12) were opportunistically available and fitted with collar monitors for the daily collection of pulse rate, respiratory rate and heart rate variability, prior to and after experimental challenge with highly pathogenic African swine fever virus. Collar monitors detected a decreased mean, and increased variability, of pulse rate and heart rate variability in pigs post-challenge, which was not detected by single daily point-in-time measurements. The incidence of abnormal pulse rate, respiratory rate and heart rate variability readings increased in pigs after infection with highly pathogenic African swine fever, with increasing abnormal readings occurring both prior to the onset of, and during, clinical disease. A preliminary non-AI algorithm utilising these data detected disease in 100%, and predicted disease onset in 67%, of infected pigs. This paper describes how health-monitoring collars can be used to improve the early detection of African swine fever disease in pigs. Additionally, it provides a potential framework for developing and using non-AI algorithms in other disease models, to enhance animal monitoring and welfare outcomes in research animals.

Castration is common in beef cattle for reproductive and management purposes, but it causes pain and stress, raising welfare concerns. The adoption of perioperative anesthesia and analgesia remains uneven in tropical low- and middle-income settings, where infrastructure, costs, and weak enforcement limit its application. This study aimed to evaluate the effects of different analgesic protocols on physiological stress markers and growth performance in beef calves raised under tropical conditions in Colombia. Forty-two intact Blanco Orejinegro (BON) calves were randomly assigned to three protocols: spermatic cord block with lidocaine plus tolfenamic acid (T1), low epidural anesthesia with lidocaine plus tolfenamic acid (T2), or tolfenamic acid alone (T3). All calves underwent closed surgical castration. Physiological and welfare indicators were assessed through pulse rate variability (PRV), hematological parameters, serum haptoglobin and cortisol concentrations, and infrared thermography (IRT) of the scrotal area at defined intervals, whereas growth performance was evaluated by average daily gain (ADG). Statistical analyses included ANOVA and nonparametric tests when assumptions were violated, with multiple comparisons corrected via Tukey’s test or Dunn’s test. No significant differences were found among the protocols for the PRV, hematological parameters, cortisol, haptoglobin, IRT, or ADG. In contrast, significant temporal effects were observed, with cortisol decreasing at 48 h, haptoglobin peaking at 48–72 h, and IRT values increasing between 24 and 48 h postcastration. These results indicate that the time after castration, rather than the analgesic protocol, was the primary driver of physiological responses.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11250-025-04832-7.

ObjectiveTo investigate whether nocturnal autonomic nervous system (ANS) activity and sleep metrics, as measured by a wearable device, can predict the occurrence of next-day migraine in patients with episodic and chronic migraine.BackgroundThe unpredictable nature of migraine episodes contributes to disease burden and limits effective application of tailored preventive strategies. Small-molecule calcitonin-gene-related peptide (CGRP) receptor antagonists, available in limited monthly quantities, are now used for both acute and preventative treatment of migraine. Consequently, improving the ability to identify days with heightened migraine risk could significantly improve migraine management and treatment outcomes.MethodsIn this prospective and observational study, adults with migraine (N = 10; 5 with chronic migraine and 5 with episodic migraine) wore the Empatica EmbracePlus®, smartwatch during sleep for a target duration of four weeks. Participants kept a headache diary recording days with no headache, non-migraine headache only, or migraine. First, group level analysis was performed using linear mixed-effects models (LMM). Next, personalized machine learning (ML) models were trained using nocturnal electrodermal activity (EDA), pulse rate variability (PRV), respiratory rate (RR), sleep duration, sleep interruptions, and awakenings to predict: (1) next-day migraine, and (2) next-day headache (both migraine and non-migraine). Performance was summarized using area under the receiver-operating and precision–recall curves (AUROC, AUPRC), sensitivity, specificity, accuracy, and precision. SHapley Additive exPlanation (SHAP) analyses identified the most influential predictors in highest performing next-day migraine and next-day headache models. Generalized Additive Models (GAM) explored nocturnal temporal dynamics of PRV and EDA.ResultsGroup level predictive performance assessed with LMMs did not reveal significant differences between ANS and sleep metrics on nights prior to no headache days, days with migraine, and days with non-migraine headache. However, individualized models using elastic-net regression, random forests, and gradient boosting machines showed modestly better-than-random AUROCs for next-day migraine prediction in 5/10 participants and next-day headache in 3/10 participants. For next-day migraine prediction models, four of five patients with episodic migraine showed better-than-random AUROCs; no patients with chronic migraine had better-than-random AUROCs. The highest-performing individualized models achieved moderate-to-good performance (AUROC 0.68 for next-day migraine and 0.81 for next-day headache). In highest performing models, SHAP analyses demonstrated sleep duration and a higher minimum PRV influenced next-day migraine and next-day headache probability, while EDA influenced next-day migraine, but not next-day headache. GAM analyses demonstrated that the first three hours after sleep onset and prior to awakening were time periods when PRV and EDA differed prior to a day with migraine or headache in these high performing models.ConclusionsOur findings indicate that applying individualized ML models to wearable-derived autonomic and sleep data may assist in the identification of heightened migraine risk and identified EDA, PRV, and sleep duration as important forecasting features. Our results provide a rationale for future studies that investigate how targeted medication and behavioral interventions on high-risk days may enhance therapeutic precision of migraine treatment and emphasize the importance of defining mechanistic subgroups of patients with migraine most likely to benefit from predictive modeling.

BackgroundInternet Gaming Disorder (IGD) has become an important emerging research topic with the popularization of Internet technology, but until now, few studies have investigated the dynamic relationship between persistent tolerance symptoms and physiological feedback. From a biomedical signal processing perspective, this gap in research development is due to the difficulty of identifying and examining the relationship between tiny, rapid fluctuations in non-linear, non-stationary physiological signals with limited time-resolved processing techniques. In this study, we propose to use high time-resolved complementary ensemble empirical pattern decomposition and normalized direct quadrature algorithm to explore the interactions between psychological characteristics and peripheral autonomic activity of individuals with IGD during a one-minute play session using the technique of instantaneous pulse rate variability (iPRV).Results26 healthy students and 20 individuals with IGD were recruited to participate in two levels of online games (G1 and G2). The IGD group reported having a fear of missing out and striving for high game standards, both reflecting the inadequacy component of tolerance symptoms. The inadequacy component was significantly positively correlated with IGD risks (r = 0.657). The IGD group also perceived G2 as significantly more challenging than G1. During the G1 stage, emotional intensity was significantly positively correlated with the time component of tolerance (r = 0.293). At the sixth minute of the G1 stage, normalized very high-frequency (r = − 0.330) and normalized low-frequency (nLF) bands iPRV (r = 0.329) were significantly negatively and positively correlated with the IGD risk, respectively. This nLF band was significantly positively correlated with the inadequacy component (r = 0.313).ConclusionThese results suggest that tolerance-related motivational components may trigger heightened emotional intensity and peripheral autonomic activation in individuals with IGD. This study provides novel insights into short-term psychophysiological dynamics underlying IGD, contributing to a deeper understanding of its persistent addictive behavior.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12938-025-01471-9.

Background: Pulse rate variability (PRV), measured through photoplethysmography (PPG), can serve as an approximate measure of cardiac autonomic regulation. Despite its increasing availability through wearable devices and its emerging clinical significance, the test-retest reliability of PRV measures remains underexplored. Methods: This study assessed PRV reliability in 832 healthy young adults (451 females; median age 28.7 years [interquartile range: 26-32]) from the Human Connectome Project – Young Adult sample. PRV was derived from cardiac signals collected during 15-minute resting-state fMRI sessions using pulse oximetry, with two sessions each day for two days. We extracted 90 distinct PRV metrics from time, frequency, and nonlinear domains using a python package NeuroKit2. Reliability across sessions and within individual days was measured using intraclass correlation coefficients (ICCs), which evaluate both consistency and absolute agreement. Results: Of 832 participants, 625 (72.5%) self-identified as white, and 482 (58.1%) completed at least a college education. The proportion of PRV measures with excellent reliability (ICC ≥0.8) was 54.4% and 41.1% on days 1 and 2, respectively, with only 28.9% demonstrating excellent reliability across both days. Specifically, 9 (28.9%) PRV measures in the time domain (Figure 1), 2 (22.2%) in the frequency domain (Figure 2), and 17 (30.4%) in the nonlinear domain showed excellent reliability (Figure 3). Measures with particularly high reliability (i.e., ICC ≥0.9) included mean (meanNN) and 80th percentile (Prc80NN) of inter-pulse intervals, percentages of successive pulse interval differences greater than 20ms (pNN20), and heart rate fragmentation (IALS). Common PRV measures such as SDNN (ICC=0.52, 95% CI: 0.47, 0.58), RMSSD (ICC=0.53, 95% CI: 0.47, 0.58), LF (ICC=0.64, 95% CI: 0.60, 0.68), and HF (ICC=0.71, 95% CI: 0.67, 0.74) exhibited lower reliability than expected, whereas normalized LF (ICC=0.83, 95% CI: 0.81, 0.85) and HF (ICC=0.86, 95% CI: 0.85, 0.88) demonstrated excellent reliability. Conclusion: Approximately 30% of PRV metrics derived from PPG signals showed excellent reliability in young adults, though validation in clinical populations remains needed. Our finding suggests that these reliable metrics could be prioritized when selecting PRV parameters for autonomic nervous system monitoring in wearable device applications.