Absolute Estimation Error Research Articles

Using spatio-temporal graph neural networks to estimate fleet-wide photovoltaic performance degradation patterns.

Accurate estimation of photovoltaic (PV) system performance is crucial for determining its feasibility as a power generation technology and financial asset. PV-based energy solutions offer a viable alternative to traditional energy resources due to their superior Levelized Cost of Energy (LCOE). A significant challenge in assessing the LCOE of PV systems lies in understanding the Performance Loss Rate (PLR) for large fleets of PV systems. Estimating the PLR of PV systems becomes increasingly important in the rapidly growing PV industry. Precise PLR estimation benefits PV users by providing real-time monitoring of PV module performance, while explainable PLR estimation assists PV manufacturers in studying and enhancing the performance of their products. However, traditional PLR estimation methods based on statistical models have notable drawbacks. Firstly, they require user knowledge and decision-making. Secondly, they fail to leverage spatial coherence for fleet-level analysis. Additionally, these methods inherently assume the linearity of degradation, which is not representative of real world degradation. To overcome these challenges, we propose a novel graph deep learning-based decomposition method called the Spatio-Temporal Graph Neural Network for fleet-level PLR estimation (PV-stGNN-PLR). PV-stGNN-PLR decomposes the power timeseries data into aging and fluctuation components, utilizing the aging component to estimate PLR. PV-stGNN-PLR exploits spatial and temporal coherence to derive PLR estimation for all systems in a fleet and imposes flatness and smoothness regularization in loss function to ensure the successful disentanglement between aging and fluctuation. We have evaluated PV-stGNN-PLR on three simulated PV datasets consisting of 100 inverters from 5 sites. Experimental results show that PV-stGNN-PLR obtains a reduction of 33.9% and 35.1% on average in Mean Absolute Percent Error (MAPE) and Euclidean Distance (ED) in PLR degradation pattern estimation compared to the state-of-the-art PLR estimation methods.

Improving Thermospheric Density Predictions in Low‐Earth Orbit With Machine Learning

AbstractThermospheric density is one of the main sources of uncertainty in the estimation of satellites' position and velocity in low‐Earth orbit. This has negative consequences in several space domains, including space traffic management, collision avoidance, re‐entry predictions, orbital lifetime analysis, and space object cataloging. In this paper, we investigate the prediction accuracy of empirical density models (e.g., NRLMSISE‐00 and JB‐08) against black‐box machine learning (ML) models trained on precise orbit determination‐derived thermospheric density data (from CHAMP, GOCE, GRACE, SWARM‐A/B satellites). We show that by using the same inputs, the ML models we designed are capable of consistently improving the predictions with respect to state‐of‐the‐art empirical models by reducing the mean absolute percentage error (MAPE) in the thermospheric density estimation from the range of 40%–60% to approximately 20%. As a result of this work, we introduce Karman: an open‐source Python software package developed during this study. Karman provides functionalities to ingest and preprocess thermospheric density, solar irradiance, and geomagnetic input data for ML readiness. Additionally, it facilitates developing and training ML models on the aforementioned data and benchmarking their performance at different altitudes, geographic locations, times, and solar activity conditions. Through this contribution, we offer the scientific community a comprehensive tool for comparing and enhancing thermospheric density models using ML techniques.

Predicting Response to In-Hospital Pulmonary Rehabilitation in Individuals Recovering From Exacerbations of Chronic Obstructive Pulmonary Disease

BackgroundPredicting the response to pulmonary rehabilitation (PR) could be valuable in defining admission priorities. We aimed to investigate whether the response of individuals recovering from a COPD exacerbation (ECOPD) could be forecasted using machine learning approaches. MethodThis multicenter, retrospective study recorded data on anthropometrics, demographics, physiological characteristics, post-PR changes in six-minute walking distance test (6MWT), Medical Research Council scale for dyspnea (MRC), Barthel Index dyspnea (BId), COPD assessment test (CAT) and proportion of participants reaching the minimal clinically important difference (MCID). The ability of multivariate approaches (linear regression, quantile regression, regression trees, and conditional inference trees) in predicting changes in each outcome measure has been assessed. ResultsIndividuals with lower baseline 6MWT, as well as those with less severe airway obstruction or admitted from acute care hospitals, exhibited greater improvements in 6MWT, whereas older as well as more dyspnoeic individuals had a lower forecasted improvement. Individuals with more severe CAT and dyspnea, and lower 6MWT had a greater potential improvement in CAT. More dyspnoeic individuals were also more likely to show improvement in BId and MRC. The Mean Absolute Error estimates of change prediction were 44.70m, 3.22 points, 5.35 points, and 0.32 points for 6MWT, CAT, BId, and MRC respectively. Sensitivity and specificity in discriminating individuals reaching the MCID of outcomes ranged from 61.78% to 98.99% and from 14.00% to 71.20%, respectively. ConclusionWhile the assessed models were not entirely satisfactory, predictive equations derived from clinical practice data might help in forecasting the response to PR in individuals recovering from an ECOPD. Future larger studies will be essential to confirm the methodology, variables, and utility.

Arrhenius Equation for Calculating Viscosity in Assessing the Dilution Level of Lubricating Oil with Diesel Oil—A Case Study of SAE 30 and SAE 40 Grade Marine Lubricating Oils

This article proposes using the Arrhenius model for estimating the viscosity of a mixture of two liquids in the quantitative assessment of the physicochemical properties of lubricating oils in the context of assessing the level of dilution of lubricating oil with diesel oil. Dynamic are made of mixtures of lubricating oil and diesel oil with known concentrations of 0, 1, 2, 5, 10, 20, 50, and 100% m/m of the diesel oil content in the mixture. Mixtures of the most popular oils (viscosity classes SAE 30 and SAE 40) with diesel oil that meet the requirements of the DMX of the marine distillate fuels category are prepared and tested. Viscosity measurements are performed at 40, 50, 60, 70, 80, 90, and 100 °C temperatures. The Arrhenius model is used to estimate the viscosity of the mixtures with an assumed diesel oil content and to estimate the diesel oil concentration in the mixtures with a known measured viscosity value. In both cases, the absolute estimation error is determined, and the accuracy of the estimation is assessed against the known concentration of diesel oil in the mixture with lubricating oil and the temperature at which the viscosity is measured. The estimated concentrations of diesel oil in mixtures with lubricating oil are useful to assess the condition of the lubricating oil (for concentrations of diesel oil lower than 5% m/m). The method is proposed to be used in practice to support standard laboratory oil analysis.

Application of UAV-Borne Visible-Infared Pushbroom Imaging Hyperspectral for Rice Yield Estimation Using Feature Selection Regression Methods

Rice yield estimation is vital for enhancing food security, optimizing agricultural management, and promoting sustainable development. However, traditional satellite/aerial and ground-based/tower-based platforms face limitations in rice yield estimation, and few studies have explored the potential of UAV-borne hyperspectral remote sensing for this purpose. In this study, we employed a UAV-borne push-broom hyperspectral camera to acquire remote sensing data of rice fields during the filling stage, and the machine learning regression algorithms were applied to rice yield estimation. The research comprised three parts: hyperspectral data preprocessing, spectral feature extraction, and model construction. To begin, the preprocessing of hyperspectral data involved geometric distortion correction, relative radiometric calibration, and rice canopy mask construction. Challenges in geometric distortion correction were addressed by tracking linear features during flight and applying a single-line correction method. Additionally, the NIR reflectance threshold method was applied for rice canopy mask construction, which was subsequently utilized for average reflectance extraction. Then, spectral feature extraction was carried out to reduce multicollinearity in the hyperspectral data. Recursive feature elimination (RFE) was then employed to identify the optimal feature set for model performance. Finally, six machine learning regression models (SVR, RFR, AdaBoost, XGBoost, Ridge, and PLSR) were used for rice yield estimation, achieving significant results. PLSR showed the best R2 of 0.827 with selected features, while XGBoost had the best R2 of 0.827 with full features. In addition, the spatial distribution of absolute error in rice yield estimation was assessed. The results suggested that this UAV-borne imaging hyperspectral-based approach held great potential for crop yield estimation, not only for rice but also for other crops.

The Optimization of the STOI Algorithm Parameters in Presence of the White Gaussian Noise (WGN)

The first part of the paper describes the Short Time Objective Intelligibility - STOI algorithm, which makes an objective evaluation of intelligibility, as well as an algorithm for estimating the optimal parameters of the STOI algorithm, N and . Second part of the paper, described an experiment that examines the intelligibility of the sentences formed from the Serbian Matrix Sentence Test - SMST base using: a). subjective MOS (Mean Opinion Score) test and b). objective test STOI algorithm. Subsequently, a comparative analysis of the results were made and it was determinate an analytical formula which connecting optimal pairs (N, ). Using the results of the MOS test as a reference, a mean absolute error of estimation of the STOI algorithm was determined.

A cost-effective joint multi-parameter optical performance monitoring scheme for high baud rate mode division multiplexing system

Inter-symbol interference (ISI), mode coupling (MC) and other impairments present in high baud rate mode division multiplexing (MDM) systems significantly degrade the reliability and accuracy of optical signal transmission. Therefore, performing joint multi-parameter optical performance monitoring (OPM) is crucial to ensure the efficient operation of the system. We propose a scheme that combines asynchronous single channel sampling (ASCS) images with a multi-task residual neural network (MT-ResNet) to enable cost-effective joint OPM at intermediate nodes in a 100 GBaud MDM system. The scheme requires only one photodetector to receive the signal. Based on the distinct characteristics exhibited by ASCS images for different parameters in the MDM system, the MT-ResNet is improved to capture correlations among parameters and learn residual information between inputs and outputs. This enables high-precision and rapid simultaneous implementation of modulation format identification (MFI), MC estimation, optical signal-to-noise ratio (OSNR) estimation, and chromatic dispersion identification (CDI) in the intermediate nodes of a 100 GBaud MDM system. The effectiveness of the proposed scheme is verified using a 100 GBaud QPSK/8QAM/16QAM/32QAM/64QAM MDM simulation transmission system with five spatial modes (LP01/LP11a/LP11b/LP21a/LP21b). The results show that under the influence of different linewidth (LW) of laser and differential mode group delay (DMGD), a 100 % identification success rate can be achieved for five commonly used MFs and eight different cumulative CD values in high baud rate MDM systems, across a wide range of OSNR. Additionally, the mean absolute error (MAE) of OSNR estimation corresponding to the five MFs is 0.22 dB, 0.26 dB, 0.30 dB, 0.32 dB, and 0.36 dB, respectively. The MAE of the eight different MC intensities fluctuate between 0.015 and 0.030. These results fully demonstrate the superior performance of the proposed scheme. Furthermore, the scheme exhibits favorable tolerance to large LWs and fiber nonlinear effects. Moreover, the complexity of the scheme is accurately analyzed through floating-point operations (FLOPs) and parameters (Params), which are much less than existing typical schemes, with specific values of 14.8849 M and 1.3798 M.

Improved State-of-Charge Estimation of Lithium-Ion Battery for Electric Vehicles Using Parameter Estimation and Multi-Innovation Adaptive Robust Unscented Kalman Filter

In this study, an improved adaptive robust unscented Kalman Filter (ARUKF) is proposed for an accurate state-of-charge (SOC) estimation of battery management system (BMS) in electric vehicles (EV). The extended Kalman Filter (EKF) algorithm is first used to achieve online identification of the model parameters. Subsequently, the identified parameters obtained from the EKF are processed to obtain the SOC of the batteries using a multi-innovation adaptive robust unscented Kalman filter (MIARUKF), developed by the ARUKF based on the principle of multi-innovation. Co-estimation of parameters and SOC is ultimately achieved. The co-estimation algorithm EKF-MIARUKF uses a multi-timescale framework with model parameters estimated on a slow timescale and the SOC estimated on a fast timescale. The EKF-MIARUKF integrates the advantages of multiple Kalman filters and eliminates the disadvantages of a single Kalman filter. The proposed algorithm outperforms other algorithms in terms of accuracy because the average root mean square error (RMSE) and the mean absolute error (MAE) of the SOC estimation were the smallest under three dynamic conditions.

A Measure of Semantic Class Difference of Point Reprojection Pairs in Camera Pose Estimation

In this paper, we propose a new measure that evaluates the semantic errors of camera poses in visual odometry (VO) and visual simultaneous localization and mapping (VSLAM). Traditionally, VO/VSLAM methods have used photometric images to estimate camera poses, but they suffer from varying illumination and viewpoint changes. Thus, methods using semantic images have been an alternative to increase consistency, as semantic information has shown its robustness even in hostile environments. Our measure compares semantic classes of map point reprojection pairs between images to improve the camera pose estimation accuracy in VO/VSLAM. To evaluate the difference between semantic classes, we adopt the normalized information distance (NID) from information theory. Furthermore, we suggest a weight parameter to balance the existing error of VO/VSLAM with the semantic error introduced by our approach. Our experimental results, obtained from the VKITTI and KITTI benchmark datasets, show that the proposed semantic error measure reduces both the relative pose error (RPE) and absolute trajectory error (ATE) of camera pose estimation compared to the existing photometric image-based errors of indirect and direct VO/VSLAM.

Are Junior Residents Accurate at Predicting Fetal Weight? An Analysis of Junior Residents' Performance of Estimated Fetal Weight Using Ultrasound and Leopold's Maneuver.

Performing accurate estimated fetal weights (EFWs) is a critical skill developed in obstetrics residency training. Resident physicians are often the first to perform EFWs on obstetric patients when they enter care. Evaluating residents' accuracy in performing EFWs is crucial for assessing their achievement in residency training milestones and providing patient care. As part of an educational initiative program between 2014 and 2020, postgraduate year 1 (PGY1) and postgraduate year 2 (PGY2) residents performed EFW measurements on 10 term (>37w0d) patients using ultrasound and Leopold's maneuver and 10 preterm (>24w0d and <37w0d) patients using ultrasound. Clinical characteristics, mode of delivery, and actual birthweights (BWs) were recorded for each patient. The accuracy of these estimates was evaluated using mixed-effect regression models. Thirty-three residents, 1127 deliveries, and 1790 EFW measurements were evaluated. Overall, the percentage of residents with estimations within 10% of actual BW went up in PGY2 for Leopold's and ultrasound term births, but not for preterm ultrasound births. Maternal body mass index and actual BW were associated with absolute percentage estimation error. After adjusting for these variables, there was a statistically significant decrease in error between PGY1 and PGY2 for Leopold's method in term births; ultrasound (term and preterm) showed more modest reductions in error during PGY2. Resident physicians have accurate estimates of EFWs early in their training, beginning in their first year of residency by both Leopold's maneuver and ultrasound. Furthermore, PGY2 residents performed better than PGY1 residents for Leopold's method.

On accurate asymptotic approximations of roots for polynomial equations containing a small, but fixed parameter

&lt;p&gt;In this paper, polynomial equations with real coefficients and in one variable were considered which contained a small, positive but specified and fixed parameter $ \varepsilon_0 \neq 0 $. By using the classical asymptotic method, roots of the polynomial equations have been constructed in the literature, which were proved to be valid for sufficiently small $ \varepsilon $-values (or equivalently for $ \varepsilon \to 0 $). In this paper, it was assumed that for some or all roots of a polynomial equation, the first few terms in a Taylor or Laurent series in a small parameter depending on $ \varepsilon $ exist and can be constructed. We also assumed that at least two approximations $ x_1(\varepsilon) $ and $ x_2(\varepsilon) $ for the real roots exist and can be constructed. For a complex root, we assumed that at least two real approximations $ a_1(\varepsilon) $ and $ a_2(\varepsilon) $ for the real part of this root, and that at least two real approximations $ b_1(\varepsilon) $ and $ b_2(\varepsilon) $ for the imaginary part of this root, exist and can be constructed. Usually it was not clear whether for $ \varepsilon = \varepsilon_0 $ the approximations were valid or not. It was shown in this paper how the classical asymptotic method in combination with the bisection method could be used to prove how accurate the constructed approximations of the roots were for a given interval in $ \varepsilon $ (usually including the specified and fixed value $ \varepsilon_0 \neq 0 $). The method was illustrated by studying a polynomial equation of degree five with a small but fixed parameter $ \varepsilon_0 = 0.1 $. It was shown how (absolute and relative) error estimates for the real and imaginary parts of the roots could be obtained for all values of the small parameter in the interval $ (0, \varepsilon_0] $.&lt;/p&gt;

Deep learning model for state of health estimation of lithium batteries based on relaxation voltage

Lithium-ion batteries have been widely used in many fields due to their high energy density, long cycle life and low self-discharge rate. However, most existing SOH estimation methods require a lengthy amount of time to determine features, the acquisition of multi-dimensional raw data is challenging, and the application scenarios are limited, making SOH estimation algorithms less practical in developing battery management systems (BMS). To face the challenge, this paper develops a new battery SOH estimation method and investigates the transferability of machine learning models under multiple operating conditions using transfer learning. First, features are extracted from the 15 s relaxation voltage using tsfresh, a Python package for time series feature extraction. Subsequently, four statistical features were obtained using a feature selection method based on the Pearson correlation coefficient. Then, a Bidirectional Long Short-Term Memory (Bi-LSTM) model is established and hyperparameters are adjusted using Bayesian optimization. The root mean square error (RMSE) and the mean absolute percentage error (MAPE) of battery SOH estimation on battery data set are 1.210 % and 1.225 %, respectively. Finally, based on transfer learning, the Bi-LSTM model is applied to battery datasets under multiple operating conditions, reducing the model's dependence on original training data while ensuring model accuracy. Considering the easy accessibility of short-term relaxation battery data, the battery SOH estimation method proposed in this paper is expected to facilitate the estimation of battery SOH under various operating conditions.

Convolutional Neural Network‐Long Short‐Term Memory‐Based State of Health Estimation for Li‐Ion Batteries under Multiple Working Conditions

The state of health (SOH) for lithium‐ion batteries is an important indicator to ensure the safety and reliability of battery energy storage systems. Aiming at the difficulty of accurately estimating the SOH of lithium‐ion batteries under different working conditions, this article proposes a method based on a hybrid convolutional neural network‐long short‐term memory (CNN‐LSTM) model. First, the battery health indicators and capacity data under different operating conditions are extracted from the public dataset to form a new dataset. Second, the CNN has multiple one‐dimensional convolutional layers to improve the efficiency of feature extraction from new datasets, and the resulting features are used as inputs to the LSTM to predict SOH. Finally, the CNN‐LSTM model integrates a fully connected layer that outputs the estimation of SOH for different operating conditions. The results show that the mean absolute error of the SOH estimation results is within 2.33% and 3.01% for the same and different working conditions, respectively.

Deep Learning-Based Oyster Packaging System

With the deepening understanding of the nutritional value of oysters by consumers, oysters as high-quality seafood are gradually entering the market. Raw edible oyster production lines mainly rely on manual sorting and packaging, which hinders the improvement of oyster packaging efficiency and quality, and it is easy to cause secondary oyster pollution and cross-contamination, which results in the waste of oysters. To enhance the production efficiency, technical level, and hygiene safety of the raw aquatic products production line, this study proposes and constructs a deep learning-based oyster packaging system. The system achieves intelligence and automation of the oyster packaging production line by integrating the deep learning algorithm, machine vision technology, and mechanical arm control technology. The oyster visual perception model is established by deep learning object detection techniques to realize fast and real-time detection of oysters. Using a simple online real-time tracking (SORT) algorithm, the grasping position of the oyster can be predicted, which enables dynamic grasping. Utilizing mechanical arm control technology, an automatic oyster packaging production line was designed and constructed to realize the automated grasping and packaging of raw edible oysters, which improves the efficiency and quality of oyster packaging. System tests showed that the absolute error in oyster pose estimation was less than 7 mm, which allowed the mechanical claw to consistently grasp and transport oysters. The static grasping and packing of a single oyster took about 7.8 s, and the success rate of grasping was 94.44%. The success rate of grasping under different transportation speeds was above 68%.

AI‐based Digital Biomarkers for Mild Dementia Decision Support using Mobile Playful Exercises with High Adherence

AbstractBackgroundEarly detection of cognitive decline and monitoring of cognitive functioning in mild dementia are fundamental for timely adaptation of lifestyle and intervention strategies. The development of digital dementia biomarkers through playful exercises with high adherence rate was a key objective of the national project multimodAAL (no. FFG 875345). The results of a study on computer‐based cognitive and physical training (CCPT) in persons diagnosed with mild Alzheimer’s disease (PwAD) are presented.MethodTablet‐PC‐based intervention was applied within 6 months in Austria, engaging PwADs living at home by means of playful multimodal training and activation (n = 11; female N = 8, male N = 3; age M = 76.6 / SD = 9.2 years, MMSE score M = 21.50 / SD = 4.41). PwADs interacted with a prototypical version of the BRAINMEE app that included a suite of cognitive exercises (puzzle, pairs, text gap filling) based on audiovisual information. The playful training app was introduced and assisted by mobile care professionals with weekly visits, however, PwDs played alone between these visits.ResultPwADs applied training with high adherence, finalizing M = 72 (M = 32) digital exercises per day within the first (last) month of the study. Duration of using exercise type ‘outsider’(p = .028*) and ‘quiz’ (p = .001**), averaged about 2 week figures, each provided statistically significant correlations (Spearman) with MMSE test scores, as well as ‘spot‐the‐difference’ (p = .003**) with Trail Making Test A, ‘outsider’ (p = .005**) with Auditory Verbal Learning Test (AVLT), respectively. A neural network (Support Vector Machine, linear kernel, 11‐fold cross validation) using duration of use of ‘quiz’, ’outsider’ and ‘hearing’ (guessing animal sounds) as input data resulted in M = 2.16 absolute error in MMSE score estimation on test data.ConclusionThe work outlined within the Austrian study on digital biomarker development indicates successful steps towards daily use of cognitive assessment using highly adherent playful training. The playful training app is applied in the European project MARA (no. FFG 886427) to enable continuous estimates of MCI’s mental state over time. The app was very well accepted by both PwADs and persons with MCI. It offers with its pervasive mental assessment tool a large potential for future long‐term monitoring in dementia prevention, early detection as well as in numerous dementia care services.

Testing the accuracy of the DRNNAGE software for age estimation in a modern Greek sample

Estimation of age-at-death from human skeletal remains is fundamental in forensic anthropology as part of the construction of the biological profile of the individual under study. At the same time, skeletal age-at-death estimation in adults is problematic due to the disparity between chronological and biological age, the important inter-individual variability at the rate of skeletal aging, and inherent biases in the available methodologies (e.g., age mimicry). A recent paper proposed a method for skeletal age-at-death estimation based on multiple anatomical traits and machine learning. A software was also created, DRNNAGE, for the easy implementation of this method. The authors of that study supported that their methods have very high repeatability and reproducibility, and the mean absolute error of the age estimation was ~6 years across the entire adult age span, which is particularly high and promising. This paper tests the proposed methodology on a modern documented Greek sample of 219 adult individuals from the Athens Collection, with age-at-death from 19 to 99 years old. The sample was split into males and females as well as into individuals under and over 50 years old. We also divided the sample in 10-year intervals. First, intra- and inter-observer error was estimated in order to assess repeatability and reproducibility of the variables employed for age-at-death estimation. Then, the validity (correct classification performance) of DRNNAGE for each anatomical region individually, as well as all combined, was evaluated on each demographic separately and on the pooled sample. According to the results, some of the variables showed very low repeatability and reproducibility, thus their use should be cautious. The DRNNAGE software showed overall highly accurate age-at-death estimates for individuals older than 50 years, but poor on younger adults, with only exception the cranial sutures, which performed surprisingly well for all age groups. Overall, these results support the importance of cross-validation and the use of population-specific methods in forensic anthropology.

State-of-health estimation method for fast-charging lithium-ion batteries based on stacking ensemble sparse Gaussian process regression

Gaussian process regression (GPR) is extensively employed in lithium-ion battery state-of-health (SOH) estimation, which ensures the safe, reliable operation of electric vehicles (EVs). However, a single GPR can produce performance discrepancies across different fast-charge batteries, as well as high time consumption. Therefore, we propose an SOH estimation method for fast-charging batteries based on stacking ensemble sparse Gaussian process regression (SGPR). First, health factors are extracted in partial discharge fragments to reflect battery degradation. Then, SGPRs based on the fully independent training condition (FITC) are developed with different kernel functions as level-1 learners, and a genetic algorithm (GA) is used to optimize the parameters of the kernel function. Further, the level-2 learner integrates the features produced by the level-1 learner based on cross validation. Finally, the accuracy, robustness, and reliability of the proposed method were evaluated under various fast-charging experiments. The results show that the mean absolute error (MAE) and root mean square error (RMSE) of SOH estimation were within 1.0852% and 1.2123%, respectively, and that the average relative time consumption was reduced by 85.68% compared with stacking ensemble GPR. Thus, the proposed method has broad application prospects in processing vast datasets from numerous batteries in monitoring platforms or cloud data centers.

Machine learning framework for Inter-Beat Interval estimation using wearable Photoplethysmography sensors

Accurate estimation of Heart Rate (HR) is crucial for various health and fitness tracking applications in wearable devices. Among the existing techniques, those based on Photoplethysmography (PPG) signals have been extensively explored for continuous HR monitoring. In this paper, we introduce a new framework that utilizes machine learning methods to estimate Inter-Beat Interval (IBI) from wrist PPG signals. The primary objective of the proposed methods is to enhance the accuracy of IBI estimates while ensuring their feasibility for wearable devices, considering typical computational power, and memory constraints. This is achieved through two combined approaches. The first method involves optimizing a set of parameters associated with the interpolation of the systolic peaks. During the training stage of this method, we optimize the weights applied to the neighboring samples of the systolic peaks while minimizing the error between the fine-tuned peak positions and their corresponding ground truth. The second method is related to a supervised machine learning approach that handles the R-peak detection as a classification task. This allows us to accurately detect systolic peaks in PPG signals by classifying candidate peaks as true or false systolic peaks. To evaluate the effectiveness of our proposed methods, we conducted experiments on a dataset collected and carefully organized to this end. It consists of PPG and Electrocardiogram (ECG) signals obtained from 46 volunteers, including individuals with normal sinus rhythm, atrial fibrillation, and other non-specified cardiac arrhythmias. Comparing our framework with other state-of-the-art methods, we observed excellent performance, with lower Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) in IBI estimation. The proposed approach also achieved higher levels of Precision and Recall in terms of the detection of systolic peaks.

SVTNet: Automatic bone age assessment network based onTW3method and vision transformer

AbstractThis study aims to develop a proficient and clinically applicable algorithm that can accurately assess bone age. This algorithm is based on the principles of the Tanner‐Whitehouse 3 (TW3) integral approach, and aims to achieve efficiency, scalability, and interpretability. We developed a model for bone age prediction in children. The model was tested on a pediatric dataset from a tertiary care hospital consisting of left‐hand radiographs of children between the age of 0 and 18. Our model consists of removing the arm portion using a pre‐trained YOLO network, localizing 37 key points in the hand bone portion using a spatial configuration network, and segmenting the original image through 20 of these points to obtain 20 fixed‐size patches. Finally, each of the 20 bone images is classified by training a visual transformer (ViT) model. In this study, a hybrid network, SVTNet, was developed that incorporates visual transformers to obtain estimates of bone age in the carpal (C series) and metacarpal (RUS series) bones. The sum of the clinical TW3 scoring region scores and bone maturity scores were utilized to determine the bone age for each corresponding region. The performance of the algorithm was evaluated in terms of both training and testing by evaluating 3871 left hand X‐ray micrographs obtained from a tertiary hospital in China. The results showed that the average absolute error of bone age estimation was 0.50 years for the RUS series of bones and 0.47 years for the C series of bones. The main contribution of this study is to propose, for the first time, a ViT‐based bone age assessment method that automates the entire process of the TW3 algorithm and is clinically interpretable, with predictive accuracy comparable to that of an experienced orthopedic surgeon.

State of health estimation for lithium-ion batteries based on two-stage features extraction and gradient boosting decision tree

To ensure the stable, reliable, and safe operation of lithium-ion batteries, the state of health (SOH) serves as a crucial basis for battery safety monitoring and operational maintenance. However, the complex working environment of batteries leads to the presence of a significant amount of noise data, abnormal data, and data discontinuities, posing challenges for battery feature extraction and SOH estimation. To address this issue, this paper proposes a two-stage transformation-based feature extraction method combined with the gradient boosting decision tree (GBDT) algorithm. The two-stage transformation feature extraction method utilizes singular spectrum analysis (SSA) and fast Fourier transform (FFT) to extract time-domain and frequency-domain health features from the battery data, respectively. The GBDT ensemble learning algorithm is employed to estimate the SOH of lithium-ion batteries, leveraging the capability to effectively uncover and learn relevant features by combining multiple weak learners into a strong model. Furthermore, correlation analysis and ablation experiments are conducted to validate the effectiveness of the two-stage feature extraction method. Additionally, four sets of comparative experiments are designed to demonstrate the superior performance of the GBDT method compared to four other machine learning methods. The experimental results show that the mean absolute error (MAE) and root mean square error (RMSE) of SOH estimation using the proposed method do not exceed 3 % on the NASA and CALCE datasets.