Physiological measurement最新文献

Objective.Individual differences across subjects reduce the accuracy of physiological signal-based cuffless blood pressure (BP) estimation. However, training a personalized model with a large amount of data is impractical. This study aims to learn a personalized model using only a few labeled samples (e.g. 5 datapoints).Approach.This study introduced a two-stage training method to enhance the few-shot personalized model with self-supervised learning. In the first training stage, self-supervised learning is used to learn shared features from physiological signals across subjects. In the second stage, few-shot learning is used to adapt the model to each subject based on the pre-trained encoder from the first stage.Main result.Experiments were conducted on the PulseDB dataset. Under the 5-shot setting, the proposed method achieved mean absolute errors of 6.57 ± 6.22 mmHg and 3.66 ± 3.99 mmHg for systolic BP (SBP) and diastolic BP (DBP) estimation, respectively, when using photoplethysmogram (PPG) and electrocardiogram. Using only PPG signals, the method achieved 6.77 ± 6.43 mmHg and 3.80 ± 3.92 mmHg for SBP and DBP estimation, respectively. The proposed approach exceeded previous non-personalized and transfer learning methods. Its generalization capability was validated on two additional smaller datasets, demonstrating the generalization ability of the proposed method.Significant.Overall, the proposed method provides a new approach for few-shot personalization of cuffless BP estimation models, which is helpful for accurate and individualized BP estimation.

Objective.Physiological measurements obtained from wearable devices reflect complex autonomic nervous system dynamics that are often assumed to follow simple linear relationships, such as elevated heart rate under stress or reduced stress during exercise. This study investigates whether physiological state recognition from wearable measurements is fundamentally linear or nonlinear by examining stress, cognitive load, and physical exercise detection.Approach.A unified signal-processing and evaluation framework was applied to three publicly available Empatica E4 datasets covering structured stress induction, real-world exam stress, aerobic and anaerobic exercise, and cognitive load tasks. Standardized preprocessing, window-based feature extraction, subject-independent evaluation, leave-one-subject-out (LOSO) validation, multimodal ablation studies, and Shapley Additive Explanations (SHAP)-based interpretability analysis were conducted. Multiple linear models (logistic regression, linear support vector machine (SVM), linear discriminant analysis, and ridge classifier) were compared against nonlinear approaches, including SVM(RBF), random forest, gradient boosting, XGBoost, and LightGBM.Main results.Across all datasets, nonlinear models consistently outperformed linear baselines. Tree-based ensembles achieved 0.89-0.98 accuracy and 0.96-0.99 AUC, whereas linear models remained below 0.70-0.73 AUC. LOSO validation revealed substantial inter-individual variability, yet nonlinear models retained moderate cross-person generalization. Ablation results confirmed the importance of multimodal fusion, particularly electrodermal activity, temperature, and accelerometry. SHAP analysis revealed nonlinear and interaction-driven feature effects consistent with known autonomic mechanisms.Significance.These findings demonstrate that physiological state recognition from wearable measurements is inherently nonlinear, even when individual modalities exhibit monotonic trends. The study establishes a unified benchmark and supports the necessity of nonlinear modeling for robust, real-time wearable health-monitoring systems.

Objective: Artificial intelligence (AI) has significantly improved image reconstruction quality across various medical imaging modalities. However, its application in electrical impedance tomography (EIT) reconstruction remains limited, mainly due to the absence of comprehensive in vivo datasets that incorporate realistic anatomical geometries and conductivity distributions. This limitation constrains the development of supervised and data-driven reconstruction methods.

Approach: To address this bottleneck, we developed StructEIT, an integrated EIT modeling framework for generating anatomically and biophysically realistic EIT simulation models. The framework incorporates three key components: (1) a structure extraction module, which automatically processes human CT scans to extract body contours and organ boundaries, thereby providing high-fidelity spatial geometry for 3D finite element modeling; (2) a surface electrode attachment module, which enables flexible and accurate placement of electrodes on irregular body surfaces, supporting diverse configurations and ensuring precise definition of the electrode-tissue interface; and (3) a tissue property assignment module, which establishes frequency-dependent conductivity models for multiple organs, enabling physiologically realistic conductivity values across tissues. Main results and Significance. By bridging the gap between CT imaging and EIT, StructEIT facilitates flexible, realistic, and scalable generation of high-resolution EIT datasets. Using this this framework, we constructed Chest-EIT, a thoracic EIT simulation dataset comprising over 1,400 publicly available CT cases, with multiple electrode configurations provided for each case.

Objective.Seismocardiography (SCG) contains rich physiological information about the structure and function of the heart, providing a new dimension for early screening and dynamic monitoring of cardiovascular diseases. However, SCG is relatively weak, susceptible to severe external interference, even has strong individual differences in morphology, making it difficult for traditional algorithms to accurately detect AO, AC and other core features.Approach.Herein, Combining the time-frequency joint distribution characteristics of SCG, we improve a novel adaptive extended Kalman filter (KF) algorithm for accurate AO and AC detection. To achieve unified modeling for different individual SCG, Gaussian mixture module is used to fit the morphological template in the time-frequency domain based on the optimal estimation strategy. Then, in order to balance the estimation accuracy and computational efficiency of the nonlinear system constructed based on SCG, the extended KF is constructed by linearizing the state transition equation. Moreover, with the aim of accurately estimating the time-varying noise components in motion, the forgetting factorαkis introduced based on the residualekwith the low-pass filtering strategy to adaptively update the measurement noise covariance matrixR, thereby achieving high-quality filtering to SCG with the AO and AC area.Main results.In addition, the experiment is conducted on the open-source CEBS dataset, indicating that the proposed algorithm has better filtering effect and higher AO and AC' detection accurate on the static SCG. Furthermore, the portable hardware system is designed for collecting SCG during 6 min walk test. Meanwhile, the impedance cardiography equipment is employed to record heart rate, left ventricular ejection time and other hemodynamics parameters. Compared with these common algorithms, the proposed algorithm also has better detection performance on the SCG during exercise.Significance.In the future, the proposed algorithm will be integrated with the portable SCG hardware system designed, which is expected to be applied in the convenient diagnosis of heart diseases, the dynamic measurement of cardiovascular parameters, the dynamic blood pressure measurement without the need for wearing cuffs and more medical scene.

Objective: We investigated (i) if blood volume decompensation status (BVDS) can be trend-tracked by hemodynamic parameters, and (ii) if hemodynamic parameters capable of trend-tracking BVDS can be trend-tracked by the physio-markers derived from the physiological signals measured using wearable sensors.

Approach: In 9 pigs undergoing controlled hemorrhage and blood transfusion, we measured gold standard arterial blood pressure (BP), heart rate (HR), stroke volume (SV), and cardiac output (CO) via invasive aortic BP and flow signals. In addition, we derived non-invasive physio-markers from the electrocardiogram (ECG), photoplethysmogram (PPG), and seismocardiogram (SCG) signals measured using wearable sensors. Then, we determined the best hemodynamic parameters to trend-track BVDS by comparing their correlation with BVDS. Finally, we investigated the feasibility of trend-tracking BVDS via non-invasive physio-markers in terms of their correlation with hemodynamic parameters as well as BVDS.

Main results: SV and CO could trend-track BVDS more consistently and explainably than BP and HR during hemorrhage and blood transfusion. The physio-markers of SV (the ratio between left ventricular ejection time (LVET) and pre-ejection period (PEP): LVET/PEP and PPG amplitude: APPG) and CO (HR·LVET/PEP and HR·APPG) showed close and monotonic relationships to SV (LVET/PEP: Spearman correlation 0.96 (0.93-0.98) and Pearson correlation 0.96 (0.93-0.98)) and CO (HR·LVET/PEP: Spearman correlation 0.95 (0.91-0.97) and Pearson correlation 0.91 (0.89-0.97)), and they likewise showed close and monotonic relationships to BVDS. However, substantial inter-individual variability in the hemodynamic parameters and their physio-markers was also observed.

Significance: These findings suggest the feasibility of wearable-enabled hemodynamic monitoring during hemorrhage and blood transfusion, as well as the challenges therein.

Objective: This work aims to enable adaptive Consumer Sleep Technologies (CSTs) for sleep intervention by developing a deep learning model for sleep stage classification using wearable sensor data.

Approach: We propose an end-to-end deep learning approach leveraging Photoplethysmography (PPG) signals, commonly available in CSTs. Model performance is improved by pretraining with Electrocardiography (ECG) from the large-scale Multi-Ethnic Study of Atherosclerosis dataset (MESA) datasets. Training and evaluation are conducted with the Dataset for Real-time sleep stage EstimAtion using Multisensor wearable Technology (DREAMT) and an additional dataset comprising of 13 participants (aged 22-71 years) without prior known sleep disorders. The dataset contains combined synchronized polysomnography (PSG) and Empatica E4 wearable data, annotated with American Academy of Sleep Medicine (AASM) sleep stages.

Main results: The proposed method demonstrates sleep stage classification from minimally processed PPG signals for real-time intervention. While ECG-trained models are not directly transferable to PPG, fine-tuning significantly improves performance, achieving up to a 29% increase in multi-stage classification accuracy.

Conclusion: Pretraining with ECG and fine-tuning with PPG increases sleep stage classification for end-to-end deep learning models, exceeding previous efforts, particularly in 3-stage sleep classification.

Significance: This work contributes to sleep health by developing a sleep stage classification model for minimally processed PPG sensor data and takes a step further towards making adaptive CSTs feasible for use with wearable sensors.

Autoregulation-guided physiological targeting, using metrics such as optimal cerebral perfusion pressure (CPPopt), optimal mean arterial pressure (MAPopt), and optimal bispectral index (BISopt), has emerged as a promising strategy for improving patient outcomes in critical care and neuromonitoring. These targets, derived from the continuous assessment of cerebrovascular reactivity (CVR) indices, are increasingly being studied for their potential to individualize patient management. This review aimed to identify and characterize existing literature detailing the derivation algorithms of CPPopt, MAPopt, and BISopt, focusing on key computational parameters, methodological consistencies, and quantitative algorithm performance metrics. Following PRISMA-ScR guidelines, studies were included if they reported algorithmic details of CPPopt, MAPopt, or BISopt derivation and provided at least six of seven core technical parameters (raw data sampling frequency, CVR index preprocessing, binning, data window size for optimality curve fitting, curve fitting method, update frequency, and yield), which were extracted during data extraction. Additional data captured included patient cohort characteristics, study objective, and CVR assessment technology. 20 studies met inclusion criteria: 13 described CPPopt, 6 described MAPopt, and 2 described BISopt derivation algorithms. CPPopt algorithms predominantly used pressure reactivity index (PRx) as the CVR index, 5 mmHg binning, and second-order polynomial curve fitting, with frequent minute-by-minute updates and multi-window averaging. MAPopt algorithms primarily used near-infrared spectroscopy (NIRS)-derived indices such as hemoglobin volume index and cerebral oximetry index (COx), while BISopt studies combined electroencephalogram (EEG) monitoring with PRx or COx. Algorithmic yield ranged from 45.6% to 100%, depending on preprocessing strategy and curve-fitting quality. Based on the existing literature, we found CPPopt derivation remains the most mature and widely studied algorithm, while MAPopt and BISopt are emerging modalities with growing interest. Despite high feasibility across studies, significant methodological variability limits the comparability of found algorithms. Standardized algorithm reporting is needed to support widespread clinical adoption of autoregulation-guided physiological targets.

Objective: Thoracic electrical bioimpedance (TEB) provides non-invasive, radiation-free monitoring of breathing. The objective of this study was to evaluate a magnitude-phase representation of TEB as a geometric and descriptive framework for respiratory signals, using short-term smoking as a test perturbation rather than a primary physiological endpoint.

Approach: Twenty-eight adult smokers (17 women, 11 men) were measured immediately before and after smoking. TEB was acquired at 50 kHz using a four-electrode thoracic configuration, and tidal volume was recorded with a pneumotachometer. Changes in impedance magnitude (|ΔZ|) and phase (Δφ) were processed using mean-centering, Hanning windowing, Fourier transformation, Gaussian band filtering around the respiratory peak, and inverse reconstruction. Lissajous plots were constructed from Δ|Z|-Δφ signals, and geometric descriptors including semi-axes (δx, δy), inclination angle (θ), ellipse area (A), eccentricity (e), and baseline offsets were extracted. Paired statistical tests were applied according to data distribution, and principal component analysis (PCA) was used to organize multiple descriptors and reduce redundancy.

Main results: Univariate analyses showed no significant pre-post differences for most variables, except for a higher mean |ΔZ| amplitude in men. In PCA space, ellipse area (A) showed consistent differences between pre- and post-smoking distributions across sexes. These differences reflected changes in joint magnitude-phase dispersion rather than statistically significant physiological effects. Inclination, semi-axes, and eccentricity showed substantial overlap between conditions. PCA provided low-dimensional representations that facilitated visualization and comparison of magnitude-phase patterns. Significance. Representing TEB signals as magnitude-phase Lissajous ellipses provides an intuitive and reproducible geometric representation of breathing. Ellipse area is proposed as a composite geometric descriptor of joint magnitude-phase variability, intended for representation and comparison rather than direct physiological inference. This non-invasive and computationally simple framework uses standard hardware and may support future methodological developments in respiratory signal analysis. .

A novel bionic esophageal device was developed to assess human swallowing function and orogastric transit, aiming ultimately to improve diagnostics for dysphagia. This miniaturized, tethered device records axial pressures, orientation, and acceleration during esophageal transit, thereby providing a dynamic view of the swallowing process. In first-inhuman feasibility tests, two healthy volunteers safely swallowed the device repeatedly in seated and supine positions. The system produced transit and pressure profiles comparable to existing technologies, with prolonged transit times observed in the supine position, e.g., transit time in seated position was median 6 s (6-23) and in the supine posture median 233 s ). These findings support the potential of this bionic device for studying esophageal motility in physiological studies as well as pathological conditions in dysphagia patients, and for future translation to untethered capsule systems capable of full gastrointestinal transit analysis.

Objective. Spirometry is the clinical gold standard for pulmonary function testing, but its reliance on mouthpiece-based airflow, trained supervision, and patient effort limits its use for frequent or home-based monitoring. This study investigates a single-point time-of-flight (TOF) sensor to capture abdominal wall displacement as a non-contact surrogate for spirometric indices.Approach. Displacement signals were recorded from 31 adult volunteers during quiet breathing, vital capacity (VC), and forced VC (FVC) manoeuvres, with simultaneous spirometry as reference. A preprocessing framework with filtering, segmentation, and feature extraction was developed, and subject-specific two-point calibration mapped displacement to lung volume. TOF-derived measures were compared to spirometry using agreement analyses, with BA plots used to quantify bias and limits of agreement for key indices.Main results. TOF signals accurately reproduced volume-related parameters: tidal volume, VC, and maximal voluntary ventilation agreed well with spirometry after calibration, with mean differences within clinically acceptable ranges. Estimation of the FEV₁/FVC ratio showed greater variability. After exclusion of one artifactual TOF measurement, BA analysis showed a small positive bias (∼+0.05) with limits of agreement of approximately -0.1 to +0.2. All TOF-derived ratios exceeded the clinical threshold of 0.7, supporting correct classification of normal ventilatory function in this cohort.Significance. These results indicate that although single-point TOF sensing cannot replace spirometry, it offers a non-contact, subject-specific calibration-minimal method for estimating pulmonary function, with promising applications in longitudinal monitoring, telehealth, and early screening.