Objective
Diabetic retinopathy (DR) is the leading cause of preventable blindness in adults and poses significant challenges in low- and middle-income regions due to limited access to skilled clinicians and diagnostic facilities. Automated screening solutions using artificial intelligence (AI) have emerged as an efficient alternative, achieving high diagnostic accuracy. However, these solutions are often developed using data from specific populations obtained using relatively expensive high-end devices. This study addresses the potential scope limitations by evaluating the AI-based screening tool retina.help across eight datasets representing diverse populations, imaging modalities, and geographic regions.

Approach
The datasets include both public and private sources, with images captured using tabletop and handheld fundus cameras. Key performance metrics for detecting binary referable DR - sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC) - were calculated on an image-by-image basis. 

Main results
Retina.help demonstrated high accuracy on tabletop images, achieving AUROC values of 0.97 on the BRSET and DeepDRiD datasets. Handheld device performance was more variable, with AUROC ranging from 0.88 (Filipino dataset) to 0.99 (Finnish dataset). Sensitivity declined with increased retinal pigmentation, as evidenced by lower values for datasets from Tanzania (62.2%) and Brazil (76.7%) compared to Finland (89.9%). Images from handheld devices often yielded lower sensitivity due to challenges related to low-contrast images. Nonetheless, retina.help generalized well across diverse datasets, showcasing its robustness.

Significance
The study highlights the impact of imaging equipment, demographics, and image quality on diagnostic performance. These findings underscore the need for benchmarking AI-based DR screening tools using standardized datasets that encompass diverse populations and imaging conditions. Such evaluations can guide the development of equitable, reliable and robust screening solutions.
.
Heart rate variability (HRV) analysis is a commonly used method, not only for assessing cardiac health but also for detecting stress and emotional arousal. However, analyzing heart rate recordings from very short (less than 1 minute) dynamic scenarios presents challenges due to their short duration, measurement artifacts, and the presence of trends. This article presents two approaches for analyzing such heart rate data. Firstly, the most common HRV metrics are reviewed, and their potential suitability is evaluated through literature review and simulated data analysis. Secondly, a novel functional approach is introduced. Both approaches are assessed and compared using real-world dataset from a dynamic simulated attack scenario. Among the selected traditional HRV metrics, RMSSD and SD1 were found to be the most suitable for analyzing ultra-short-term dynamic scenarios. The functional approach yielded similar conclusions to the traditional approach while providing additional possibilities for further analysis and deeper interpretation of results. Finally, the advantages, disadvantages, and potential applications of both approaches are discussed.
Objective.The accuracy of psychological stress detection hinges upon the precision with which models capture complex and variable physiological responses. Current research primarily employs deep learning techniques to model multimodal physiological responses, aiming to enhance detection performance by leveraging complementary features across modalities. However, existing deep learning approaches face challenges in modeling long-range dependencies and exhibit suboptimal robustness in cross-modal fusion.Approach.To address this, this paper proposes ResoMamba-LLM, a multimodal stress detection framework integrating Mamba and large language models (LLM). Specifically, ResoMamba-LLM employs dual-channel Mamba encoders to process chest and wrist physiological signals respectively, supplemented by a frequency domain branch to capture periodic rhythms. It further utilizes lightweight reprogramming techniques to map physiological features onto the semantic space of the LLM, thereby stimulating its latent contextual reasoning capabilities.Main results.Achieving accuracies of 95.83% and 81.65% on the WESAD and EmoWear datasets, respectively, experimental results conclusively demonstrate ResoMamba-LLM's robust long-dependency modeling capabilities for multimodal physiological signals and its robust cross-modal fusion abilities.Significance.This provides a novel solution for precise, interpretable psychological stress detection.
OBJECTIVE:To develop an artifact-management method suitable for low-density, dry-electrode EEG recorded from a wearable eyewear platform, enabling artifact detection, classification, and correction while avoiding unnecessary processing of clean neural activity. APPROACH:A gated two-stage pipeline is proposed in which a lightweight gate classifier first identifies artifact-contaminated EEG windows and selectively triggers further processing only when needed. The complete pipeline combines: (i) a time-domain feature-based detector, (ii) a convolutional-recurrent network for three-class artifact discrimination (eye blinks, horizontal eye movements, facial movements), and (iii) a UNet-based denoising autoencoder. MAIN RESULTS:Evaluated on the eyewear EEG dataset (29 subjects) using a leave-one-subject-out scheme, the gate classifier achieved a balanced accuracy of 0.89. The artifact classifier reached a balanced accuracy of 0.84 with a median macro-F1 of 0.83. The denoiser achieved a median normalized RMSE lower than 0.20 and a spectral cosine similarity above 0.84. In the end-to-end evaluation, the pipeline achieved a median normalized RMSE of 0.19 and a spectral cosine similarity of 0.89. Evaluation on an independent external benchmark showed that the proposed pipeline maintained a slightly lower gate detection performance (balanced accuracy 0.82) and achieved higher two-class classification performance. The denoising performance was comparable to that obtained on the dataset collected for this study, with absolute median differences not exceeding 0.04 for both normalized RMSE and spectral cosine similarity. SIGNIFICANCE:These results support the effectiveness of the proposed approach for artifact management in wearable eyewear EEG, while maintaining a design compatible with resource-constrained wearable implementations.
Objective.Atrial fibrillation (AF) burden has become an increasingly important endpoint in long-duration rhythm monitoring, but reliable burden estimation requires more than accurate AF detection alone. In particular, when burden is derived by aggregating predicted AF probabilities over time, probability calibration may directly affect burden validity under external dataset shift.Approach.This study developed an interpretable-interval feature model for AF detection and evaluated it using record-wise cross-validation on a development cohort and independent cross-dataset external validation on public Holter electrocardiographic databases. Window-level performance was assessed using the area under the receiver operating characteristic curve (ROC-AUC), area under the precision-recall curve (PR-AUC), Brier score, expected calibration error (ECE), and calibration intercept and calibration slope. Recording-level AF burden was estimated using both probability-based and hard-label aggregation and evaluated using mean absolute error (MAE) and agreement analyses.Main results.The model showed high discrimination in both development and external evaluation, with external ROC-AUC ofand PR-AUC of. However, external calibration deteriorated despite preserved ranking performance, with Brier score of, ECE(15) of, calibration intercept of, and calibration slope of. In the external cohort, probability-based burden estimation preserved strong association with reference burden but showed weaker raw agreement than hard-label aggregation, with MAE ofversus, consistent with systematic probability underprediction. Repeated external recalibration across record-level splits substantially improved probability quality and probability-based burden estimation. Median probability-burden MAE decreased fromwithout recalibration toafter Platt recalibration andafter isotonic recalibration, while median ECE(15) decreased fromtoand, respectively.Significance.These findings indicate that-interval-based AF detection maintained strong ranking performance in the tested external cohort, but probability calibration should be evaluated explicitly when predicted probabilities are aggregated into AF-burden estimates.
Objective. Fetal electrocardiography (fECG) provides beat-to-beat cardiac information, offering the potential to improve the assessment of fetal response to uterine activity. However, the accuracy of fECG-derived cardiotocograms depends on signal quality. Scalp electrode recordings offer reliable fECG but are invasive and measurable only during labor. In contrast, non-invasive fECG measurements suffer from a low signal-to-noise ratio and maternal interference. To advance research on the use of non-invasive fECG, there is a need for large datasets with non-invasive multi-channel fECG recorded simultaneously with a clear reference fECG, such as from a scalp electrode.Approach. We propose a method to address this data scarcity by synthesizing multi-channel fECG signals by modifying inter-beat intervals based on a target fetal heart rate (fHR). The proposed method adjusts the duration of the TP segments between consecutive fECG complexes to effectively match the inter-beat intervals of the target fHR. The adjusted multi-channel TP segments are generated by a generative model that has learned the structure of the TP segments in our dataset. The morphology of the fetal QRS complexes remains unchanged. The fidelity of the synthesized signals is evaluated using signal quality indices (SQIs) and fHR estimates.Main results. Most SQIs computed on the input and synthesized signals exhibited strong correlation (), with low mean absolute differences. The error in the standard deviation of the estimated fHR and that of the target fHR wasBPM. Morphology-based SQIs exhibited strong correlations (), validating the preservation of fetal QRS complexes.Significance. Our approach generates realistic multi-channel fECG signals that match the target fHR patterns, supporting data augmentation and algorithm development for fetal monitoring. It addresses the scarcity of multi-channel fECG measurements with a clean reference available for algorithm development. It could enhance the robustness of data-driven and source separation methods.
Objective.Chagas disease, a neglected tropical disease (NTD) with significant cardiovascular impact, remains underdiagnosed in resource-limited regions. Electrocardiogram (ECG) screening offers a low-cost tool for detecting cardiac involvement, yet algorithm development is challenged by label noise, data scarcity, and the latent nature of infection. This study proposes a robust ECG-based screening framework that explicitly addresses these constraints.Approach.We introduce aReliability-Aware Hierarchical Learningstrategy that calibrates supervision according to data provenance, prioritizing serology-confirmed labels over noisy self-reports. To mitigate data scarcity, we compare a specialized convolutional neural network (CNN) trained from scratch with a transfer learning approach based on a Spatio-Temporal ECG foundation Model (FM). Performance is evaluated across varying data scales, and the representation structure is analyzed to interpret model behavior.Main results.On the official hidden test set of the George B. Moody PhysioNet/Computing in Cardiology Challenge 2025, our approach achieved a Challenge Score of 0.163. We observe that while the specialized CNN performs competitively in data-rich regimes, the FM exhibits superior robustness in extreme low-resource settings. Furthermore, performance reaches a plateau imposed by underlying disease physiology. Bimodal score distributions suggest that models distinguish established cardiomyopathy from indeterminate infection, which remains electrophysiologically indistinguishable from healthy controls.Significance.These findings clarify both the potential and intrinsic limits of ECG-based AI screening for NTD-associated cardiac involvement. Reliability-aware supervision and data-efficient transfer learning provide a practical framework toward scalable and clinically meaningful ECG screening systems in resource-constrained environments.
Objective.Heterogeneous treatment effect (HTE) estimation is essential for understanding individual differences in physiological responses and supporting personalized healthcare. However, existing non-parametric HTE estimation methods often rely on complex partition strategies and may have limited interpretability when applied to physiological measurements with continuous variations and non-uniform data distributions. This study aims to develop an adaptive and interpretable framework for HTE estimation in physiological measurement systems.Approach.We propose a sliding-window-based double machine learning framework (Slide-DML) for non-parametric HTE estimation. Slide-DML adaptively constructs quasi-homogeneous local windows based on treatment effect variation and estimates local linear HTE within each selected window. The local estimates are subsequently aggregated using adaptive weighting to obtain a smooth global HTE function while preserving interpretability.Main results.The performance of Slide-DML was evaluated using synthetic data, semi-synthetic clinical data, and real physiological measurements. In synthetic experiments, Slide-DML achieved a mean squared error (MSE) of 0.006, outperforming existing machine learning-based HTE estimation methods. In semi-synthetic experiments, Slide-DML achieved a root MSE of 3.526, demonstrating superior performance compared with both machine learning-based and deep learning-based approaches. Experiments on real photoplethysmogram and electrocardiogram data further showed that the estimated treatment effect curves were consistent with established cardiovascular knowledge and provided improved interpretability.Significance.Slide-DML provides an effective and interpretable approach for estimating HTE in physiological measurement systems. By capturing continuous variations in physiological states, the proposed framework may facilitate individualized analysis of cardiovascular responses and support personalized healthcare applications, such as cuffless blood pressure monitoring using wearable physiological devices.
Objective.Pulmonary function tests are central to assessing ventilatory impairment but are often limited by accessibility, repeatability, and patient compliance, particularly in community and longitudinal monitoring settings. This study investigates whether acoustic features of voluntary coughs can serve as low burden surrogate signals reflecting trends in spirometry related indices of ventilatory impairment.Approach.We propose a dual output ensemble learning framework that integrates deterministic regression and probabilistic interval estimation to jointly model spirometry related parameters from cough acoustics and basic demographic information. Fifty eight cough derived acoustic features and four biometric variables were analysed using an optimized extreme gradient boosting (XGBoost) model for point estimation, coupled with natural gradient boosting (NGBoost) for uncertainty quantification. A hybrid Bayesian-Whale Optimization strategy was employed for robust hyperparameter tuning.Main results.In a cohort of 278 subjects with heterogeneous respiratory conditions, the proposed framework achieved root mean square errors of 0.20 l for forced vital capacity (FVC) and 0.27 l for forced expiratory volume in one second (FEV1), with coefficients of determination up to 0.94 and 0.87, respectively. Interval estimates provided uncertainty bounds consistent with model reliability, with the actual prediction interval coverage probabilities reaching 84.0% for FVC and 77.8% for FEV1. Performance for the composite FEV1/FVC ratio was comparatively lower, reflecting the increased variability of ratio based metrics.Significance.These findings suggest that cough acoustics encode physiologically relevant information associated with ventilatory impairment and can support proxy trend modelling of spirometry related indices, rather than direct spirometry replacement. The proposed framework demonstrates the feasibility of cough based physiological assessment for low cost respiratory screening and longitudinal monitoring, particularly where conventional spirometry is impractical. Larger multi centre studies are warranted to further evaluate generalizability and clinical integration.
BACKGROUND AND OBJECTIVE:Automated respiratory sound classification based on deep learning is critical for enhancing diagnostic precision and efficiency in respiratory disease. Research on respiratory sound classification for adults is extensive, whereas work on paediatric respiratory sounds is still scarce. Compared to adults, paediatric respiratory sounds typically contain more high-frequency components, and this inherent difference presents unique challenges for models. APPROACH:In response to this challenge, we proposed a novel architecture, which integrated the local feature extraction module with the global context model. Specifically, we designed the Mel_Grouper module to serve as a front-end with the purpose of enhancing local pathological representations. The output was fed into the Transformer-based Mel_Encoder to fuse global context. MAIN RESULTS:Experimental results on the SJTU Paediatric Respiratory Sound (SPRSound) dataset demonstrate that our method achieves state-of-the-art (sota) performance. Regarding the four subtasks of the SPRSound dataset, it outperforms the previous best results by 3.37%, 3.06%, 6.83% and 5.36%, respectively. To better reflect the real clinical conditions, we further performed experiments on the real-world paediatric respiratory sound dataset. SIGNIFICANCE:These results demonstrate the method's significant performance in paediatric respiratory sound classification. The code is available at: https://github.com/shufei2580/SSAST/tree/master.
Objective.To evaluate thermal recovery after controlled cooling using dynamic infrared thermography and to determine whether the recovery process can be described using distinct temporal components associated with different heat transfer dynamics.Approach.A controlled cooling protocol was applied to the plantar region of the foot, followed by thermal recovery monitoring using a radiometric infrared camera. Temperature evolution was analyzed using a two-time-constant exponential model, whereτ1andτ2represent the slow and fast recovery components, respectively. A frame-by-frame radiometric calibration procedure was implemented to ensure measurement stability. The distributions ofτ1andτ2were evaluated under different cooling durations and recovery windows using non-parametric statistics.Main results.Thermal recovery was consistently described by two-time constants that capture distinct temporal regimes. The fast component (τ2) exhibited relatively compact distributions, particularly in the short recovery window, indicating stable and repeatable behavior. In contrast, the slow component (τ1) showed larger variability, especially for longer recovery windows. This variability suggests that the slow component (τ1) is more sensitive to extended thermal dynamics and inter-subject variability than to measurement noise. The results also showed that longer cooling durations increased the relative contribution of the slow recovery component (τ1), revealing multi-scale thermal recovery behavior.Significance.These findings demonstrate that thermal recovery is not a single-scale process but a multi-time dynamic phenomenon in which slow and fast recovery components provide complementary information. By explicitly separating these temporal components, the proposed framework enables a more detailed characterization of thermal dynamics than single-parameter approaches. As a methodological proof-of-concept, this work supports the use of dynamic infrared thermography as a quantitative tool for physiological measurement and establishes a reproducible basis for future clinical studies.
Objective.Pulse-to-pulse intervals obtained from continuous non-invasive blood pressure (CNBP) signals can be used to derive pulse rate variability (PRV). CNBP signals are susceptible to noise and motion artifacts, which can reduce the reliability of PRV and its agreement with heart rate variability (HRV) derived from electrocardiography (ECG). In this study, we applied a pulse-wise, unsupervised quality assessment method for CNBP signals based on a convolutional autoencoder integrating waveform morphology with pulse-level metadata. The aim was to improve the agreement between PRV and HRV metrics.Approach.Two datasets of CNBP and ECG recordings were analyzed: long baseline measurements (15 volunteers, average 40 min) and short recordings with postural changes (53 volunteers; 5 min squat-stand). A convolutional autoencoder was used for pulse-wise quality assessment. HRV metrics were derived from ECG, and PRV metrics were derived from CNBP before and after quality-based pulse selection (PRV). The analysis included standard time-domain indices (SDNN, RMSSD), frequency-domain measures, and nonlinear Poincaré plot descriptors (SD1, SD2). The agreement between HRV and PRV or PRVwas evaluated by Bland-Altman analysis.Main results.PRV systematically overestimated variability compared to HRV. During resting conditions, RMSSD and SDNN were lower for HRV (47.720.6 ms and 71.427.4 ms) than for PRV (61.726.7 ms,= 0.005 and 77.830.0 ms,= 0.003, respectively). After applying quality-based pulse selection, bias decreased fromtoms for RMSSD and fromtoms for SDNN, improving PRV-HRV agreement. During postural changes, PRV also overestimated HRV, but the effect of quality-based pulse selection was less pronounced.Significance.The proposed approach enhances CNBP signal quality without requiring manual labeling, making it suitable for real-world applications. Improved agreement between PRV and HRV may have important clinical implications, particularly when ECG acquisition is impractical or unavailable.
Background. Cerebral autoregulation (CA) is estimated by assessing the association between variations of mean cerebral blood velocity (MCBv) and mean arterial pressure (MAP). Recently, regional oxygen saturation (rSO2) has been tested as an alternative to MCBv for CA estimation.Objective. We propose a correlation-based method iterating the computation of the Pearson correlation coefficientrover short windows of MAP, MCBv and rSO2and testing on an individual basis the significance of the positive MCBv-MAP and rSO2-MAP association.Analysis. The rejection of the null hypothesis of zero or negative correlation was performed using a fast approach based on the classical t-test applied to the Pearson correlation coefficient and via a time-consuming approach based on surrogate series generation. The median ofrcomputed over segments rejecting the null hypothesis and their percentage was computed in 53 patients (age: 62 ± 11 years, 39 males, 14 females) scheduled for cardiac surgery acquired before (BASAL) and after induction of propofol-based general anesthesia (ANESTH). The method was compared to more traditional time-domain techniques.Main results.The percentage of positively correlated sequences and theirrdecreased during ANESTH and this result was valid regardless of the method utilized to reject the null hypothesis and the use of MCBv or rSO2. Since markers computed using MCBv and rSO2were uncorrelated, two approaches might unveil different CA aspects. Only results based on MCBv were significantly correlated with traditional time-domain indexes.Significance.The method should be considered for applications of CA monitoring in practical settings.
Objective.This study aimed to quantify cardiopulmonary physiological modulation in patients with hypertension (HTN) and ischaemic heart disease (IHD) using electrocardiogram (ECG) and respiration signals recorded before and after a 7 d Panchakarma-based integrative residential programme.Approach.Simultaneous 10 min ECG and respiration recordings were acquired before and after the programme. A predefined signal-quality pipeline was applied before graph construction. ECG-derived cardiac dynamics were based on normal-to-normal (NN) intervals after artefact and ectopy screening, while respiration cycles were screened for physiologically plausible breath intervals. From each acceptable recording, a clean 300 s segment was selected. NN-derived heart rate (HR) and respiration-rate series were interpolated to a common 4 Hz time base, detrended and standardised. Natural visibility graphs (VGs), weighted VGs and a two-layer multilayer VG (MLVG) were then constructed using synchronous HR-respiration interlayer coupling. Graph index complexity (GIC), average path length, assortativity and pulse-respiration quotient PRQ-derived features were analysed using paired Wilcoxon signed-rank tests with Hodges-Lehmann paired differences and false discovery rate correction.Main results.After quality control, alignment and graph-eligibility screening, paired graph-feature analysis included 20 HTN and 26 IHD subjects. The alignment procedure produced complete HR-respiration time series with negligible missing fractions. Post-program recordings showed measurable cardiopulmonary modulation. In HTN, PRQ decreased with false-discovery-rate-supported significance, while weighted GIC features showed descriptive post-program increases. In IHD, unweighted respiration-VG and MLVG GIC increased descriptively, whereas weighted and path-based metrics showed heterogeneous changes. Interlayer surrogate validation further quantified whether identity-coupled MLVG complexity exceeded shifted-coupling null networks.Significance.The proposed framework provides a reproducible ECG-respiration network-physiology workflow combining NN-based quality control, respiration-cycle screening, common-time-base alignment, unimodal and multilayer VG modelling, PRQ analysis and surrogate validation. The findings support objective quantification of post-program cardiopulmonary modulation while avoiding causal attribution to Panchakarma alone.
Objective.While core body temperature in humans remains relatively stable within a narrow range, peripheral skin temperature fluctuates dynamically. This variability reflects complex interactions between heat loss (e.g. via vasodilation) and heat gain (e.g. through metabolic activity), all modulated by an integrated thermoregulatory network. Emerging evidence suggests that skin temperature variability (STV) may offer insight into the integrity of thermoregulatory, vascular, and autonomic control systems. This systematic review evaluates the prognostic utility of STV as a dynamic physiological marker across diverse clinical settings.Method.The review followed PRISMA guidelines, with comprehensive searches conducted in Ovid MEDLINE, EMBASE, and AMED up to June 2026. Studies examining STV in relation to prognosis were eligible, and methodological quality was assessed using the QualSyst and QUIPS tools. This study used a narrative evidence synthesis guided by a logic model.Results.Of 41 papers screened, 10 met the inclusion criteria. The included studies identified a range of analytic methods used to quantify STV. Overall, the prognostic value of STV appeared to depend on both the timescale of temperature fluctuations and the underlying disease context. In life-threatening conditions such as sepsis, multiple organ failure and decompensated cirrhosis, reduced short-term variability or loss of complexity predicted higher mortality, potentially reflecting severe autonomic or microvascular regulation. At longer timescales, attenuated 24-hour circadian amplitude of temperature time-series in healthy individuals predicted future cardiometabolic diseases, indicating circadian dysregulation associated with increased all-cause mortality risk.Discussion.STV shows promise as a non-invasive physiological marker of prognosis, but current evidence is limited by methodological heterogeneity, small clinical cohorts, and inconsistent confounder control. Standardised measurement protocols, agreed analytical metrics, multicentre validation, and large-scale comparison with established risk scores are needed to enable clinical implementation.
Objective.The widespread adoption of wearable ECG devices has driven an explosive growth of long-term, multi-lead ECG data. However, clinical analysis and model development remain constrained by the excessive cost of data annotation, a limitation particularly pronounced in conventional deep learning methods that rely on labeled information.Approach.To address the problem of label scarcity in wearable ECG measurement scenarios, this paper proposes a self-supervised learning method for ECG based on intervention-based attribution alignment, aiming to improve the model's analytical performance on sparsely labeled ECG signals using unlabeled data. Our method integrates an intervention-inspired operation into the contrastive learning framework: it extracts compact representations from unlabeled data on one hand, and employs feature intervention to learn view-invariant mechanisms on the other, thereby reducing dependency on labeled data for target tasks.Main results.Experimental results on four public ECG datasets demonstrate that the proposed method achieves performance on par with state-of-the-art approaches in typical tasks such as arrhythmia classification, while utilizing only a fraction of the annotated samples.Significance.The findings indicate that our method effectively accommodates the multi-scenario monitoring characteristics of wearable devices, offering a practical solution to alleviate ECG data annotation burdens and advance the efficient analysis of long-term ECG signals.
Objective.To improve deep learning-based cuffless blood pressure (BP) estimation from photoplethysmography (PPG) for continuous, non-invasive monitoring when cuff measurements are impractical, by assessing sampling frequency, systolic BP (SBP) and diastolic BP (DBP) derivation, respiratory rate (RR) effects, and errors at BP extremes.Approach.We analysed 205 850 ten-second paired PPG-arterial BP segments from the medical information mart for intensive care waveform database. A convolutional model with a multiresolution U-Net translated PPG to arterial BP waveforms. Inputs at 125, 62.5, and 31.25 Hz were compared. SBP and DBP were derived by minimum-maximum extraction or by averaging the top 4 systolic peaks and bottom 4 diastolic troughs per segment. Performance was summarized using mean absolute error (MAE).Main results.Inputs at 62.5 Hz reduced MAE versus 125 Hz (SBP 5.24 vs 6.34 mmHg; DBP 2.87 vs 3.27 mmHg), whereas 31.25 Hz increased error (SBP 16.09 mmHg; DBP 9.10 mmHg). Multi-peak averaging outperformed minimum-maximum extraction. MAE varied across RRs, with generally lower errors observed at moderate RRs, and higher errors observed at lower and higher SBP ranges.Significance.Sampling frequency, SBP/DBP derivation, and RR shape error patterns in PPG-only cuffless BP estimation, with higher errors at low and high SBP. These results inform device design for bedside and ambulatory monitoring. However, further validation using subject-level datasets is required to assess clinical applicability and compliance with established validation standards.
Objective.Low tidal volume, limited driving pressure and positive end expiratory pressure (PEEP) are well-accepted lung-protected ventilation strategies. The study aimed to compare driving pressure targeted strategy (PTS) and low tidal volume + PEEP strategy (VPS) in term of ventilation heterogeneity under mechanical ventilation via mask during anesthetic induction.Approach. Patients undergoing anesthesia induction were randomly assign to either PTS or VPS groups. An inspiration pressure of 15 cm H2O and zero PEEP was used in PTS group. A tidal volume of 8 ml kg-1predicted body weight and a PEEP of 5 cm H2O was used in VPS group. The regional lung ventilation was monitored by electrical impedance tomography (EIT). Expiratory tidal volume (TVe), driving pressure, hemodynamics and PaO2/FiO2, and EIT-derived ventilation homogeneity (global inhomogeneity index,GI), the ratio of left/right lung ventilation distribution were calculated.Main results.Finally, 30 patients with PTS and 25 with VPS were analyzed. Compared with spontaneous ventilation, mechanical ventilation via mask caused significant changes in theGI(0.40 ± 0.04 vs 0.48 ± 0.08,P< 0.05). When compared with PTS, VPS was associated with improvedGI(0.52 ± 0.09 vs 0.42 ± 0.03, PTS vs VPS,P< 0.001, power value 0.955), lowerTVe (625.3 ± 192.7 vs 429.1 ± 48.8 ml,P< 0.001), lower driving pressure (15 ± 0 vs 10.5 ± 1.5 cm H2O,P< 0.001), and slightly higher but not statistically different PaO2/FiO2(370.7 ± 90.1 vs 391.1 ± 48.9 mm Hg),P= 0.380).Significance. Mask mechanical ventilation with VPS may improve the lung ventilation inhomogeneity during anesthetic induction when compared with PTS.
Objective.Adjustment of positive end-expiratory pressure (PEEP) is part of lung-protective ventilation in patients with acute respiratory distress syndrome (ARDS) and can be performed invasively using an esophageal catheter (Talmoret al2008New Engl. J. Med.3592095-104). The objective of this proof-of-concept study was to investigate whether the transpulmonary pressure (TPP) of 0 mbar, at which alveolar collapse occurs, can be estimated non-invasively using electrical impedance tomography (EIT).Approach.ARDS was induced in 14 pigs by repeated lung lavage and a 2 h period of ventilator-induced lung injury ventilation. The TPP was calculated as the difference between end-expiratory airway pressure and the corresponding esophageal pressure. During a descending pressure-controlled ramp maneuver from 50 to 0 mbar, pressure-flow curves at the pixel levels of the candidate dependent lung region were generated and regional deflation parameters calculated.Main results.During the descending pressure ramp the beginning of alveolar collapse could be determined at approximately 25 mbar in the dependent region of the EIT images based on regional characteristic points in both lung conditions. A strong correlation was found (r= 0.929,B= 1.005,R2= 0.862,p= 0.007) in healthy lungs and an intermediate correlation (r= 0.817,B= 0.894,R2= 0.668,p= 0.025) in ARDS between the second characteristic point and the PEEP according to Talmor, corresponding to a TPP of 0.Significance.The findings of the study have demonstrated that EIT can identify regional characteristic points that correlate with regional TPP changes in the dependent lung region. This knowledge can be used to optimize a lung protective ventilation strategy by titrating TPP guided PEEP non-invasively.
Background.During emergency transport, clinical assessment and vital signs (VS) may lack the sensitivity to identify traumatic brain injury (TBI) and identify specific TBI subtypes which may have implications for triaging and timely delivery of life-saving interventions.Objective.To evaluate the ability of machine learning (ML) algorithms to identify the presence of TBI with or without specific important concomitant clinical phenotypes including shock, coagulopathy and polytrauma during air transport to a trauma center.Methods.We identified a cohort of consecutive trauma patients aged 18-65 transported from the scene of injury via helicopter to an urban academic trauma center and collected prehospital clinical data and continuous VSs. We used ElasticNet (regularized regression) and XGBoost (gradient boosting), comparing three variable sets: clinical variables only, continuous physiologic monitoring data only, and combined clinical and physiological data, to develop four predictive models: (1) presence/absence of TBI, (2) mild vs moderate-severe TBI, (3) presence/absence of polytrauma in moderate-severe TBI, (4) presence/absence of coagulopathy in TBI, and (5) presence/absence of shock in TBI.Results.1025 patients (median age 38, interquartile range (IQR): 27-53; 70% male; median Glasgow coma scale 15 (IQR: 13-15) were identified. Across all predictive models, ML algorithms exhibited good predictive discrimination, with area under the receiver operator curve of 0.79 (0.75-0.84), 0.79 (0.74-0.83), 0.89 (0.85-0.92), 0.77 (0.67-0.86), and 0.78 (0.72-0.84) for TBI, TBI severity, polytrauma, coagulopathy, and shock, respectively. Clinical data best predicted TBI severity and polytrauma, while physiologic data improved prediction of shock and coagulopathy.Conclusions.ML algorithms integrating clinical and continuous physiological monitoring can improve identification of TBI and concomitant clinical phenotypes during prehospital transport.
