Methods to estimate aspects of physical activity and sedentary behavior from high-frequency wrist accelerometer measurements

This investigation developed models to estimate aspects of physical activity and sedentary behavior from three-axis high-frequency wrist-worn accelerometer data. The models were developed and tested on 20 participants ( n = 10 males, n = 10 females, mean age = 24.1, mean body mass index = 23.9), who wore an ActiGraph GT3X+ accelerometer on their dominant wrist and an ActiGraph GT3X on the hip while performing a variety of scripted activities. Energy expenditure was concurrently measured by a portable indirect calorimetry system. Those calibration data were then used to develop and assess both machine-learning and simpler models with fewer unknown parameters (linear regression and decision trees) to estimate metabolic equivalent scores (METs) and to classify activity intensity, sedentary time, and locomotion time. The wrist models, applied to 15-s windows, estimated METs [random forest: root mean squared error (rSME) = 1.21 METs, hip: rMSE = 1.67 METs] and activity intensity (random forest: 75% correct, hip: 60% correct) better than a previously developed model that used counts per minute measured at the hip. In a separate set of comparisons, the simpler decision trees classified activity intensity (random forest: 75% correct, tree: 74% correct), sedentary time (random forest: 96% correct, decision tree: 97% correct), and locomotion time (random forest: 99% correct, decision tree: 96% correct) nearly as well or better than the machine-learning approaches. Preliminary investigation of the models' performance on two free-living people suggests that they may work well outside of controlled conditions.

Download Full-text

Calibrating Wrist-Worn Accelerometers for Physical Activity Assessment in Preschoolers: Machine Learning Approaches (Preprint)

10.2196/preprints.16727 ◽

2019 ◽

Author(s):

Shiyu Li ◽

Jeffrey T Howard ◽

Erica T Sosa ◽

Alberto Cordova ◽

Deborah Parra-Medina ◽

...

Keyword(s):

Physical Activity ◽

Machine Learning ◽

Cluster Analysis ◽

Sedentary Behavior ◽

Classification Accuracy ◽

Characteristic Curve ◽

Learning Approaches ◽

Cut Points ◽

Activity Intensity ◽

Vector Magnitude

BACKGROUND Physical activity (PA) level is associated with multiple health benefits during early childhood. However, inconsistency in the methods for quantification of PA levels among preschoolers remains a problem. OBJECTIVE This study aimed to develop PA intensity cut points for wrist-worn accelerometers by using machine learning (ML) approaches to assess PA in preschoolers. METHODS Wrist- and hip-derived acceleration data were collected simultaneously from 34 preschoolers on 3 consecutive preschool days. Two supervised ML models, receiver operating characteristic curve (ROC) and ordinal logistic regression (OLR), and one unsupervised ML model, k-means cluster analysis, were applied to establish wrist-worn accelerometer vector magnitude (VM) cut points to classify accelerometer counts into sedentary behavior, light PA (LPA), moderate PA (MPA), and vigorous PA (VPA). Physical activity intensity levels identified by hip-worn accelerometer VM cut points were used as reference to train the supervised ML models. Vector magnitude counts were classified by intensity based on three newly established wrist methods and the hip reference to examine classification accuracy. Daily estimates of PA were compared to the hip-reference criterion. RESULTS In total, 3600 epochs with matched hip- and wrist-worn accelerometer VM counts were analyzed. All ML approaches performed differently on developing PA intensity cut points for wrist-worn accelerometers. Among the three ML models, k-means cluster analysis derived the following cut points: ≤2556 counts per minute (cpm) for sedentary behavior, 2557-7064 cpm for LPA, 7065-14532 cpm for MPA, and ≥14533 cpm for VPA; in addition, k-means cluster analysis had the highest classification accuracy, with more than 70% of the total epochs being classified into the correct PA categories, as examined by the hip reference. Additionally, k-means cut points exhibited the most accurate estimates on sedentary behavior, LPA, and VPA as the hip reference. None of the three wrist methods were able to accurately assess MPA. CONCLUSIONS This study demonstrates the potential of ML approaches in establishing cut points for wrist-worn accelerometers to assess PA in preschoolers. However, the findings from this study warrant additional validation studies.

Download Full-text

Machine Learning Classifiers Based Classification For IRIS Recognition

Qubahan Academic Journal ◽

10.48161/qaj.v1n2a48 ◽

2021 ◽

Vol 1 (2) ◽

pp. 106-118

Author(s):

Bahzad Taha Chicho ◽

Adnan Mohsin Abdulazeez ◽

Diyar Qader Zeebaree ◽

Dilovan Assad Zebari

Keyword(s):

Machine Learning ◽

Random Forest ◽

Decision Tree ◽

Iris Recognition ◽

Nearest Neighbors ◽

Comparison Analysis ◽

K Nearest Neighbors ◽

Applied Machine Learning ◽

Data Objects ◽

Better Than

Classification is the most widely applied machine learning problem today, with implementations in face recognition, flower classification, clustering, and other fields. The goal of this paper is to organize and identify a set of data objects. The study employs K-nearest neighbors, decision tree (j48), and random forest algorithms, and then compares their performance using the IRIS dataset. The results of the comparison analysis showed that the K-nearest neighbors outperformed the other classifiers. Also, the random forest classifier worked better than the decision tree (j48). Finally, the best result obtained by this study is 100% and there is no error rate for the classifier that was obtained.

Download Full-text

Calibrating Wrist-Worn Accelerometers for Physical Activity Assessment in Preschoolers: Machine Learning Approaches

JMIR Formative Research ◽

10.2196/16727 ◽

2020 ◽

Vol 4 (8) ◽

pp. e16727

Author(s):

Shiyu Li ◽

Jeffrey T Howard ◽

Erica T Sosa ◽

Alberto Cordova ◽

Deborah Parra-Medina ◽

...

Keyword(s):

Physical Activity ◽

Machine Learning ◽

Cluster Analysis ◽

Sedentary Behavior ◽

Classification Accuracy ◽

Characteristic Curve ◽

Learning Approaches ◽

Cut Points ◽

Activity Intensity ◽

Vector Magnitude

Background Physical activity (PA) level is associated with multiple health benefits during early childhood. However, inconsistency in the methods for quantification of PA levels among preschoolers remains a problem. Objective This study aimed to develop PA intensity cut points for wrist-worn accelerometers by using machine learning (ML) approaches to assess PA in preschoolers. Methods Wrist- and hip-derived acceleration data were collected simultaneously from 34 preschoolers on 3 consecutive preschool days. Two supervised ML models, receiver operating characteristic curve (ROC) and ordinal logistic regression (OLR), and one unsupervised ML model, k-means cluster analysis, were applied to establish wrist-worn accelerometer vector magnitude (VM) cut points to classify accelerometer counts into sedentary behavior, light PA (LPA), moderate PA (MPA), and vigorous PA (VPA). Physical activity intensity levels identified by hip-worn accelerometer VM cut points were used as reference to train the supervised ML models. Vector magnitude counts were classified by intensity based on three newly established wrist methods and the hip reference to examine classification accuracy. Daily estimates of PA were compared to the hip-reference criterion. Results In total, 3600 epochs with matched hip- and wrist-worn accelerometer VM counts were analyzed. All ML approaches performed differently on developing PA intensity cut points for wrist-worn accelerometers. Among the three ML models, k-means cluster analysis derived the following cut points: ≤2556 counts per minute (cpm) for sedentary behavior, 2557-7064 cpm for LPA, 7065-14532 cpm for MPA, and ≥14533 cpm for VPA; in addition, k-means cluster analysis had the highest classification accuracy, with more than 70% of the total epochs being classified into the correct PA categories, as examined by the hip reference. Additionally, k-means cut points exhibited the most accurate estimates on sedentary behavior, LPA, and VPA as the hip reference. None of the three wrist methods were able to accurately assess MPA. Conclusions This study demonstrates the potential of ML approaches in establishing cut points for wrist-worn accelerometers to assess PA in preschoolers. However, the findings from this study warrant additional validation studies.

Download Full-text

Comparisons of physical activity and sedentary behavior between owners and non-owners of commercial wearable devices

Perspectives in Public Health ◽

10.1177/1757913921989389 ◽

2021 ◽

Vol 141 (2) ◽

pp. 89-96

Author(s):

Hsin-Yen Yen ◽

Hao-Yun Huang

Keyword(s):

Physical Activity ◽

Sleep Quality ◽

Sedentary Behavior ◽

Sedentary Time ◽

Short Form ◽

Wearable Devices ◽

Wearable Device ◽

World Health ◽

World Health Organization Quality ◽

Quality Of Life Scale

Aims: Wearable devices are a new strategy for promoting physical activity in a free-living condition that utilizes self-monitoring, self-awareness, and self-determination. The main purpose of this study was to explore health benefits of commercial wearable devices by comparing physical activity, sedentary time, sleep quality, and other health outcomes between individuals who used and those that did not use commercial wearable devices. Methods: The research design was a cross-sectional study using an Internet survey in Taiwan. Self-administered questionnaires included the International Physical Activity Questionnaire–Short Form, Pittsburgh Sleep Quality Index, Health-Promoting Lifestyle Profile, and World Health Organization Quality-of-Life Scale. Results: In total, 781 participants were recruited, including 50% who were users of wearable devices and 50% non-users in the most recent 3 months. Primary outcomes revealed that wearable device users had significantly higher self-reported walking, moderate physical activity, and total physical activity, and significantly lower sedentary time than non-users. Wearable device users had significantly better sleep quality than non-users. Conclusion: Wearable devices inspire users’ motivation, engagement, and interest in physical activity through habit formation. Wearable devices are recommended to increase physical activity and decrease sedentary behavior for promoting good health.

Download Full-text

Machine Learning in Aging: An Example of Developing Prediction Models for Serious Fall Injury in Older Adults

Innovation in Aging ◽

10.1093/geroni/igaa057.859 ◽

2020 ◽

Vol 4 (Supplement_1) ◽

pp. 268-269

Author(s):

Jaime Speiser ◽

Kathryn Callahan ◽

Jason Fanning ◽

Thomas Gill ◽

Anne Newman ◽

...

Keyword(s):

Machine Learning ◽

Older Adults ◽

Random Forest ◽

Decision Tree ◽

Prediction Models ◽

Receiver Operating Curve ◽

Learning Methods ◽

Life Study ◽

Fall Injury ◽

Machine Learning Methods

Abstract Advances in computational algorithms and the availability of large datasets with clinically relevant characteristics provide an opportunity to develop machine learning prediction models to aid in diagnosis, prognosis, and treatment of older adults. Some studies have employed machine learning methods for prediction modeling, but skepticism of these methods remains due to lack of reproducibility and difficulty understanding the complex algorithms behind models. We aim to provide an overview of two common machine learning methods: decision tree and random forest. We focus on these methods because they provide a high degree of interpretability. We discuss the underlying algorithms of decision tree and random forest methods and present a tutorial for developing prediction models for serious fall injury using data from the Lifestyle Interventions and Independence for Elders (LIFE) study. Decision tree is a machine learning method that produces a model resembling a flow chart. Random forest consists of a collection of many decision trees whose results are aggregated. In the tutorial example, we discuss evaluation metrics and interpretation for these models. Illustrated in data from the LIFE study, prediction models for serious fall injury were moderate at best (area under the receiver operating curve of 0.54 for decision tree and 0.66 for random forest). Machine learning methods may offer improved performance compared to traditional models for modeling outcomes in aging, but their use should be justified and output should be carefully described. Models should be assessed by clinical experts to ensure compatibility with clinical practice.

Download Full-text

Modified Decision Tree Technique for Ransomware Detection at Runtime through API Calls

Scientific Programming ◽

10.1155/2020/8845833 ◽

2020 ◽

Vol 2020 ◽

pp. 1-10

Author(s):

Faizan Ullah ◽

Qaisar Javaid ◽

Abdu Salam ◽

Masood Ahmad ◽

Nadeem Sarwar ◽

...

Keyword(s):

Machine Learning ◽

Random Forest ◽

Decision Tree ◽

Feature Vector ◽

Machine Learning Algorithms ◽

The Novel ◽

Proposed Model ◽

Testing Accuracy ◽

Financial Losses

Ransomware (RW) is a distinctive variety of malware that encrypts the files or locks the user’s system by keeping and taking their files hostage, which leads to huge financial losses to users. In this article, we propose a new model that extracts the novel features from the RW dataset and performs classification of the RW and benign files. The proposed model can detect a large number of RW from various families at runtime and scan the network, registry activities, and file system throughout the execution. API-call series was reutilized to represent the behavior-based features of RW. The technique extracts fourteen-feature vector at runtime and analyzes it by applying online machine learning algorithms to predict the RW. To validate the effectiveness and scalability, we test 78550 recent malign and benign RW and compare with the random forest and AdaBoost, and the testing accuracy is extended at 99.56%.

Download Full-text

Leading by Example: Association Between Mother and Child Objectively Measured Physical Activity and Sedentary Behavior

Pediatric Exercise Science ◽

10.1123/pes.2020-0058 ◽

2020 ◽

pp. 1-12

Author(s):

Brad R. Julius ◽

Amy M.J. O’Shea ◽

Shelby L. Francis ◽

Kathleen F. Janz ◽

Helena Laroche

Keyword(s):

Physical Activity ◽

Sedentary Behavior ◽

Low Income ◽

Multivariate Regression ◽

Sedentary Time ◽

Vigorous Physical Activity ◽

Mother And Child ◽

Low Income Mothers ◽

Objectively Measured ◽

The Relationship

Purpose: The authors examined the relationship between mother and child activity. Methods: The authors compared moderate–vigorous physical activity (MVPA) and sedentary time of low-income mothers with obesity and their 6- to 12-year-old children on week (WD) and weekend (WE) days. A total of 196 mother–child pairs wore accelerometers simultaneously for a week. Mothers completed questionnaires. Spearman correlation and multivariate regression were used. Results: WE MVPA (accelerometry) was significantly correlated between mothers with children aged 6–7 (rs = .35) and daughters (rs = .27). Self-reported maternal PA time spent with one of their children was significantly correlated with the WE MVPA of all children (rs = .21) and children aged 8–10 (rs = .22) and with the WD MVPA of all children (rs = .15), children aged 8–10 (rs = .23), aged 11–12 (rs = .52), and daughters (rs = .37), and inversely correlated to the WD sedentary time of all children (rs = −.21), children aged 8–10 (rs = −.30), aged 11–12 (rs = −.34), daughters (rs = −.26), and sons (rs = −.22). In multivariate regression, significant associations were identified between reported child–mother PA time together and child MVPA and sedentary time (accelerometry). Conclusions: Mothers may influence the PA levels of their children with the strongest associations found in children aged 6–7 and daughters. Mother–child coparticipation in PA may lead to increased child MVPA and decreased sedentary behavior.

Download Full-text

Abstract 035: Chronic Mild Sleep Restriction Leads To Increased Sedentary Time And Reductions In Physical Activity

Circulation ◽

10.1161/circ.143.suppl_1.035 ◽

2021 ◽

Vol 143 (Suppl_1) ◽

Author(s):

Faris M Zuraikat ◽

Samantha Scaccia ◽

Keith M Diaz ◽

Justin Cochran ◽

Arindam RoyChoudhury ◽

...

Keyword(s):

Physical Activity ◽

Sedentary Behavior ◽

Sedentary Time ◽

Sleep Restriction ◽

Positive Energy ◽

Full Sample ◽

Men And Women ◽

Short Sleep ◽

Wrist Actigraphy ◽

Positive Energy Balance

Background: Sleeping less than 7 h per night is a risk factor for positive energy balance and weight gain. While the effect of short sleep on energy intake has been extensively studied, its influence on physical activity (PA), a key determinant of energy expenditure, is not well characterized. To date, no study has evaluated sedentary and PA patterns in response to chronic mild short sleep, which is experienced by up to one-third of US adults. Hypothesis: Sedentary behavior will be higher and PA (light intensity [LIPA] and moderate-vigorous intensity [MVPA]) will be lower during 6 wk of mild sleep restriction (SR) relative to maintenance of adequate sleep (AS). Methods: Data are presented from 76 participants (age: 34.2±12.4 y; BMI: 25.6±3.4 kg/m 2 ; n=55 women) from two randomized crossover trials with identical sleep interventions. Men and women with adequate habitual sleep duration ≥7 h/ night underwent two 6-wk sleep conditions, AS and SR, separated by a 6-wk washout period. During AS, participants were instructed to maintain average nightly bed and wake times determined from 2 wk screening with wrist-actigraphy and sleep logs. In SR, total sleep time was curtailed by 1.5 h per night by delaying bedtimes. Nightly sleep diaries and 24-h wrist actigraphy confirmed adherence to the protocol, which was verified weekly. Daily wrist actigraphy data were used to determine time spent in sedentary behavior and PA. Linear mixed models were used to test whether sleep condition (SR vs AS) influenced sedentary behavior or PA, adjusting for time in bed. Results: Across the full sample, sedentary time was significantly greater in SR than in AS (39.8±13.6 min/d, P<0.01). Similar results were observed in analyses stratified by sex; compared to AS, in SR, sedentary time was 53.0±16.5 min/d higher in women (P=0.001) with a trend towards significance in men (20.3±11.3, P=0.07). Although a slight increase in LIPA over 6 wk was observed in SR relative to AS in the full sample (2.9±0.8 min/d, P<0.001) and in men (3.7±1.2 min/d, P<0.01), overall, time spent in LIPA across weeks was significantly lower in SR relative to AS. This main effect of SR on LIPA was observed in the full sample (SR vs AS: -44.6±3.3 min/d, P<0.0001) and separately in women (SR vs AS: -38.2±10.5 min/d, P<0.001) and men (SR vs AS: -9.4±4.6 min/d, P=0.04). In men only, the slope of change in MVPA over 6 wk differed slightly in SR vs AS (2.6±1.1, P=0.02). However, across weeks, time in MVPA was significantly lower in SR compared to HS (-12.4±4.2 min/d, P=0.003). Conclusions: We provide some of the first evidence of an adverse impact of chronic short sleep on PA patterns in men and women. Greater sedentary time and lower PA levels can promote positive energy balance and may underlie associations of short sleep with risk for cardiometabolic diseases. Results further highlight the importance of achieving adequate sleep to promote cardiovascular health.

Download Full-text

Reconstruction of high frequency methane peaks from measurements of metal oxide low-cost sensors using machine learning

10.5194/egusphere-egu21-16027 ◽

2021 ◽

Author(s):

Rodrigo Rivera Martinez ◽

Diego Santaren ◽

Olivier Laurent ◽

Ford Cropley ◽

Cecile Mallet ◽

...

Keyword(s):

Machine Learning ◽

Random Forest ◽

Metal Oxide ◽

High Frequency ◽

Polynomial Regression ◽

Low Cost ◽

Point Sources ◽

Baseline Correction ◽

High Concentration ◽

Industrial Facilities

Deploying a dense network of sensors around emitting industrial facilities allows to detect and quantify possible CH4&#160;leaks and monitor the emissions continuously. Designing such a monitoring network with highly precise instruments is limited by the elevated cost of instruments, requirements of power consumption and maintenance. Low cost and low power metal oxide sensor could come handy to be an alternative to deploy this kind of network at a fraction of the cost with satisfactory quality of measurements for such applications.Recent studies have tested Metal Oxide Sensors (MOx) on natural and controlled conditions to measure atmospheric methane concentrations and showed a fair agreement with high precision instruments, such as those from Cavity Ring Down Spectrometers (CRDS). Such results open perspectives regarding the potential of MOx to be employed as an alternative to measure and quantify CH4 emissions on industrial facilities. However, such sensors are known to drift with time, to be highly sensitive to water vapor mole fraction, have a poor selectivity with several known cross-sensitivities to other species and present significant sensitivity environmental factors like temperature and pressure. Different approaches for the derivation of CH4 mole fractions from the MOx signal and ancillary parameter measurements have been employed to overcome these problems, from traditional approaches like linear or multilinear regressions to machine learning (ANN, SVM or Random Forest).Most studies were focused on the derivation of ambient CH4 concentrations under different conditions, but few tests assessed the performance of these sensors to capture CH4 variations at high frequency, with peaks of elevated concentrations, which corresponds well with the signal observed from point sources in industrial sites presenting leakage and isolated methane emission. We conducted a continuous controlled experiment over four months (from November 2019 to February 2020) in which three types of MOx Sensors from Figaro&#174; measured high frequency CH4 peaks with concentrations varying between atmospheric background levels up to 24 ppm at LSCE, Saclay, France. We develop a calibration strategy including a two-step baseline correction and compared different approaches to reconstruct CH4 spikes such as linear, multilinear and polynomial regression, and ANN and random forest algorithms. We found that baseline correction in the pre-processing stage improved the reconstruction of CH4 concentrations in the spikes. The random forest models performed better than other methods achieving a mean RMSE = 0.25 ppm when reconstructing peaks amplitude over windows of 4 days. In addition, we conducted tests to determine the minimum amount of data required to train successful models for predicting CH4 spikes, and the needed frequency of re-calibration / re-training under these controlled circumstances. We concluded that for a target RMSE <= 0.3 ppm at a measurement frequency of 5s, 4 days of training are required, and a recalibration / re-training is recommended every 30 days.Our study presents a new approach to process and reconstruct observations from low cost CH4 sensors and highlights its potential to quantify high concentration releases in industrial facilities.

Download Full-text

Subepithelial neutrophil infiltration as a predictor of the surgical outcome of chronic rhinosinusitis with nasal polyps

Rhinology Journal ◽

10.4193/rhin20.373 ◽

2020 ◽

Vol 0 (0) ◽

pp. 0-0

Author(s):

D-K. Kim ◽

H-S. Lim ◽

K.M. Eun ◽

Y. Seo ◽

J.K. Kim ◽

...

Keyword(s):

Machine Learning ◽

Random Forest ◽

Decision Tree ◽

Chronic Rhinosinusitis ◽

Surgical Outcomes ◽

Nasal Polyps ◽

Human Neutrophil Elastase ◽

Immunofluorescence Analysis ◽

Double Positive ◽

Ki 67

BACKGROUND: Neutrophils present as major inflammatory cells in refractory chronic rhinosinusitis with nasal polyps (CRSwNP), regardless of the endotype. However, their role in the pathophysiology of CRSwNP remains poorly understood. We investigated factors predicting the surgical outcomes of CRSwNP patients with focus on neutrophilic localization. METHODS: We employed machine-learning methods such as the decision tree and random forest models to predict the surgical outcomes of CRSwNP. Immunofluorescence analysis was conducted to detect human neutrophil elastase (HNE), Bcl-2, and Ki-67 in NP tissues. We counted the immunofluorescence-positive cells and divided them into three groups based on the infiltrated area, namely, epithelial, subepithelial, and perivascular groups. RESULTS: On machine learning, the decision tree algorithm demonstrated that the number of subepithelial HNE-positive cells, Lund-Mackay (LM) scores, and endotype (eosinophilic or non-eosinophilic) were the most important predictors of surgical outcomes in CRSwNP patients. Additionally, the random forest algorithm showed that, after ranking the mean decrease in the Gini index or the accuracy of each factor, the top three ranking factors associated with surgical outcomes were the LM score, age, and number of subepithelial HNE-positive cells. In terms of cellular proliferation, immunofluorescence analysis revealed that Ki-67/HNE-double positive and Bcl-2/HNE-double positive cells were significantly increased in the subepithelial area in refractory CRSwNP. CONCLUSION: Our machine-learning approach and immunofluorescence analysis demonstrated that subepithelial neutrophils in NP tissues had a high expression of Ki-67 and could serve as a cellular biomarker for predicting surgical outcomes in CRSwNP patients.

Download Full-text