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

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.

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

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.

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

2015 ◽  
Vol 119 (4) ◽  
pp. 396-403 ◽  
Author(s):  
John Staudenmayer ◽  
Shai He ◽  
Amanda Hickey ◽  
Jeffer Sasaki ◽  
Patty Freedson
Keyword(s):  
Physical Activity ◽  
Machine Learning ◽  
Random Forest ◽  
Decision Tree ◽  
Sedentary Behavior ◽  
High Frequency ◽  
Sedentary Time ◽  
Correct Decision ◽  
Activity Intensity ◽  
Better Than

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.

scholarly journals Impact of ActiGraph Sampling Rate and Intermonitor Comparability on Measures of Physical Activity in Adults

2021 ◽  
pp. 1-11
Author(s):  
Kimberly A. Clevenger ◽  
Jan Christian Brønd ◽  
Daniel Arvidsson ◽  
Alexander H.K. Montoye ◽  
Kelly A. Mackintosh ◽  
...  
Keyword(s):  
Physical Activity ◽  
Sampling Rate ◽  
Mean Amplitude ◽  
Moderate Intensity ◽  
Equivalence Testing ◽  
Cut Points ◽  
Activity Intensity ◽  
Moderate Intensity Activity ◽  
Mean Differences ◽  
Vector Magnitude

Background: ActiGraph is a commonly used, research-grade accelerometer brand, but there is little information regarding intermonitor comparability of newer models. In addition, while sampling rate has been shown to influence accelerometer metrics, its influence on measures of free-living physical activity has not been directly studied. Purpose: To examine differences in physical activity metrics due to intermonitor variability and chosen sampling rate. Methods: Adults (n = 20) wore two hip-worn ActiGraph wGT3X-BT monitors for 1 week, with one accelerometer sampling at 30 Hz and the other at 100 Hz, which was downsampled to 30 Hz. Activity intensity was classified using vector magnitude, Euclidean Norm Minus One (ENMO), and mean amplitude deviation (MAD) cut points. Equivalence testing compared outcomes. Results: There was a lack of intermonitor equivalence for ENMO, time in sedentary/light- or moderate-intensity activity according to ENMO cut points, and time in moderate-intensity activity according to MAD cut points. Between sampling rates, differences existed for time in moderate-intensity activity according to vector magnitude, ENMO, and MAD cut points, and time in sedentary/light-intensity activity according to ENMO cut points. While mean differences were small (0.1–1.7 percentage points), this would equate to differences in moderate-to vigorous-intensity activity over a 10-hr wear day of 3.6 (MAD) to 10.8 (ENMO) min/day for intermonitor comparisons or 3.6 (vector magnitude) to 5.4 (ENMO) min/day for sampling rate. Conclusions: Epoch-level intermonitor differences were larger than differences due to sampling rate, but both may impact outcomes such as time spent in each activity intensity. ENMO was the least comparable metric between monitors or sampling rates.

scholarly journals Associations between objectively measured physical activity, sedentary behaviour and time in bed among 75+ community-dwelling Danish older adults

2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Li-Tang Tsai ◽  
Eleanor Boyle ◽  
Jan C. Brønd ◽  
Gry Kock ◽  
Mathias Skjødt ◽  
...  
Keyword(s):  
Physical Activity ◽  
Older Adults ◽  
Sedentary Behavior ◽  
Sedentary Behaviour ◽  
Sleep Duration ◽  
Activity Level ◽  
Time In Bed ◽  
Objectively Measured ◽  
Objectively Measured Physical Activity ◽  
Vector Magnitude

Abstract Background Older adults are recommended to sleep 7–8 h/day. Time in bed (TIB) differs from sleep duration and includes also the time of lying in bed without sleeping. Long TIB (≥9 h) are associated with self-reported sedentary behavior, but the association between objectively measured physical activity, sedentary behavior and TIB is unknown. Methods This study was based on cross-sectional analysis of the Healthy Ageing Network of Competence (HANC Study). Physical activity and sedentary behaviour were measured by a tri-axial accelerometer (ActiGraph) placed on the dominant wrist for 7 days. Sedentary behavior was classified as < 2303 counts per minute (cpm) in vector magnitude and physical activity intensities were categorized, as 2303–4999 and ≥ 5000 cpm in vector magnitude. TIB was recorded in self-reported diaries. Participants were categorized as UTIB (usually having TIB 7–9 h/night: ≥80% of measurement days), STIB (sometimes having TIB 7–9 h/night: 20–79% of measurement days), and RTIB (rarely having TIB 7–9 h/night: < 20% of measurement days). Multinominal regression models were used to calculate the relative risk ratios (RRR) of being RTIB and STIB by daily levels of physical activity and SB, with UTIB as the reference group. The models were adjusted for age, sex, average daily nap length and physical function. Results Three hundred and fourty-one older adults (median age 81 (IQR 5), 62% women) were included with median TIB of 8 h 21 min (1 h 10 min)/day, physical activity level of 2054 (864) CPM with 64 (15) % of waking hours in sedentary behavior. Those with average CPM within the highest tertile had a lower RRR (0.33 (0.15–0.71), p = 0.005) for being RTIB compared to those within the lowest tertile of average CPM. Accumulating physical activity in intensities 2303–4999 and ≥ 5000 cpm/day did not affect the RRR of being RTIB. RRR of being RTIB among highly sedentary participants (≥10 h/day of sedentary behavior) more than tripled compared to those who were less sedentary (3.21 (1.50–6.88), p = 0.003). Conclusions For older adults, being physically active and less sedentary was associated with being in bed for 7–9 h/night for most nights (≥80%). Future longitudinal studies are warranted to explore the causal relationship sbetween physical activity and sleep duration.

scholarly journals Phenotyping Women Based on Dietary Macronutrients, Physical Activity, and Body Weight Using Machine Learning Tools

Nutrients ◽  
2019 ◽  
Vol 11 (7) ◽  
pp. 1681 ◽  
Author(s):  
Ramyaa Ramyaa ◽  
Omid Hosseini ◽  
Giri P. Krishnan ◽  
Sridevi Krishnan
Keyword(s):  
Physical Activity ◽  
Machine Learning ◽  
Cluster Analysis ◽  
Body Weight ◽  
Energy Balance ◽  
Numerical Data ◽  
Support Vector ◽  
Learning Tools ◽  
K Nearest Neighbor ◽  
Numerical Predictions

Nutritional phenotyping can help achieve personalized nutrition, and machine learning tools may offer novel means to achieve phenotyping. The primary aim of this study was to use energy balance components, namely input (dietary energy intake and macronutrient composition) and output (physical activity) to predict energy stores (body weight) as a way to evaluate their ability to identify potential phenotypes based on these parameters. From the Women’s Health Initiative Observational Study (WHI OS), carbohydrates, proteins, fats, fibers, sugars, and physical activity variables, namely energy expended from mild, moderate, and vigorous intensity activity, were used to predict current body weight (both as body weight in kilograms and as a body mass index (BMI) category). Several machine learning tools were used for this prediction. Finally, cluster analysis was used to identify putative phenotypes. For the numerical predictions, the support vector machine (SVM), neural network, and k-nearest neighbor (kNN) algorithms performed modestly, with mean approximate errors (MAEs) of 6.70 kg, 6.98 kg, and 6.90 kg, respectively. For categorical prediction, SVM performed the best (54.5% accuracy), followed closely by the bagged tree ensemble and kNN algorithms. K-means cluster analysis improved prediction using numerical data, identified 10 clusters suggestive of phenotypes, with a minimum MAE of ~1.1 kg. A classifier was used to phenotype subjects into the identified clusters, with MAEs <5 kg for 15% of the test set (n = ~2000). This study highlights the challenges, limitations, and successes in using machine learning tools on self-reported data to identify determinants of energy balance.

scholarly journals Dynamic Detection of Delayed Cerebral Ischemia Using Machine Learning

2020 ◽  
Author(s):  
Murad Megjhani ◽  
Kalijah Terilli ◽  
Ayham Alkhachroum ◽  
David J. Roh ◽  
Sachin Agarwal ◽  
...  
Keyword(s):  
Machine Learning ◽  
Cerebral Ischemia ◽  
Characteristic Curve ◽  
Delayed Cerebral Ischemia ◽  
Risk Scores ◽  
Support Vector ◽  
Model Parameters ◽  
Learning Approaches ◽  
Physiologic Data ◽  
Over Time

AbstractObjectiveTo develop a machine learning based tool, using routine vital signs, to assess delayed cerebral ischemia (DCI) risk over time.MethodsIn this retrospective analysis, physiologic data for 540 consecutive acute subarachnoid hemorrhage patients were collected and annotated as part of a prospective observational cohort study between May 2006 and December 2014. Patients were excluded if (i) no physiologic data was available, (ii) they expired prior to the DCI onset window (< post bleed day 3) or (iii) early angiographic vasospasm was detected on admitting angiogram. DCI was prospectively labeled by consensus of treating physicians. Occurrence of DCI was classified using various machine learning approaches including logistic regression, random forest, support vector machine (linear and kernel), and an ensemble classifier, trained on vitals and subject characteristic features. Hourly risk scores were generated as the posterior probability at time t. We performed five-fold nested cross validation to tune the model parameters and to report the accuracy. All classifiers were evaluated for good discrimination using the area under the receiver operating characteristic curve (AU-ROC) and confusion matrices.ResultsOf 310 patients included in our final analysis, 101 (32.6%) patients developed DCI. We achieved maximal classification of 0.81 [0.75-0.82] AU-ROC. We also predicted 74.7 % of all DCI events 12 hours before typical clinical detection with a ratio of 3 true alerts for every 2 false alerts.ConclusionA data-driven machine learning based detection tool offered hourly assessments of DCI risk and incorporated new physiologic information over time.

scholarly journals Machine learning approaches revealed metabolic signatures of incident chronic kidney disease in persons with pre-and type 2 diabetes

2020 ◽  
Author(s):  
Ada Admin ◽  
Jialing Huang ◽  
Cornelia Huth ◽  
Marcela Covic ◽  
Martina Troll ◽  
...  
Keyword(s):  
Machine Learning ◽  
Type 2 Diabetes ◽  
Chronic Kidney Disease ◽  
Kidney Disease ◽  
Characteristic Curve ◽  
Predictive Performance ◽  
Population Based ◽  
Learning Approaches ◽  
Clinical Variables

Early and precise identification of individuals with pre-diabetes and type 2 diabetes (T2D) at risk of progressing to chronic kidney disease (CKD) is essential to prevent complications of diabetes. Here, we identify and evaluate prospective metabolite biomarkers and the best set of predictors of CKD in the longitudinal, population-based Cooperative Health Research in the Region of Augsburg (KORA) cohort by targeted metabolomics and machine learning approaches. Out of 125 targeted metabolites, sphingomyelin (SM) C18:1 and phosphatidylcholine diacyl (PC aa) C38:0 were identified as candidate metabolite biomarkers of incident CKD specifically in hyperglycemic individuals followed during 6.5 years. Sets of predictors for incident CKD developed from 125 metabolites and 14 clinical variables showed highly stable performances in all three machine learning approaches and outperformed the currently established clinical algorithm for CKD. The two metabolites in combination with five clinical variables were identified as the best set of predictors and their predictive performance yielded a mean area value under the receiver operating characteristic curve of 0.857. The inclusion of metabolite variables in the clinical prediction of future CKD may thus improve the risk prediction in persons with pre- and T2D. The metabolite link with hyperglycemia-related early kidney dysfunction warrants further investigation.

scholarly journals Revisiting the MotionWatch8©: Calibrating Cut-Points for Measuring Physical Activity and Sedentary Behavior Among Adults With Stroke

2019 ◽  
Vol 11 ◽  
Author(s):  
Ryan S. Falck ◽  
John R. Best ◽  
Michael C. R. Li ◽  
Janice J. Eng ◽  
Teresa Liu-Ambrose
Keyword(s):  
Physical Activity ◽  
Sedentary Behavior ◽  
Cut Points

Research Tracker 6 Accelerometer Calibration and Validation in Comparison to GENEActiv, ActiGraph, and Gas Analysis in Young Adults

2019 ◽  
Vol 2 (3) ◽  
pp. 176-187
Author(s):  
Emma L. J. Eyre ◽  
Jason Tallis ◽  
Susie Wilson ◽  
Lee Wilde ◽  
Liam Akhurst ◽  
...  
Keyword(s):  
Physical Activity ◽  
Young Adults ◽  
Convergent Validity ◽  
Complex Analysis ◽  
Axial Vector ◽  
Gas Analysis ◽  
Moderate Activity ◽  
Sedentary Behaviors ◽  
Cut Points ◽  
Activity Intensity

Background: The ability to objectively assess physical activity and inactivity in free living individuals is important in understanding activity patterns and the dose response relationship with health. Currently, a large number of research tools exist, but little evidence has examined the validity/utility of the Research Tracker 6 (RT6) monitor. Questions remain in regard to the best placements, positions, and cut-points in young adults to determine activity intensity across a range of activities. This study sought to address this gap in young adults. The study aims were 1) to examine criterion validity of RT6 in comparison to breath-by-breath gas analysis; 2) convergent validity of RT6 in comparison to ActiGraph and GENEActiv; 3) development of RT6 tri-axial vector magnitude cut-points to classify physical activity at different intensities (i.e., for sedentary, moderate, and vigorous); 4) to compare the generated cut-points of the RT6 in comparison to other tools. Methods: Following ethics approval and informed consent, 31 young adults (age = 22±3 years: BMI = 23±3 kg/m2) undertook five modes of physical activity/sedentary behaviors while wearing three different accelerometers at hip and wrist locations (ActiGraph GT9X Link, GENEActiv, RT6). Expired gas was sampled during the five activities (MetaMax 3B). Correlational analysis assessed the relationship between accelerometer devices and METs/VO2. Receiver Operating Characteristic Curves analysis were used to calculate area under the curve and define cut-points for physical activity intensities. Results: The RT6 demonstrated criterion and convergent validity (r = 0.662–0.966, P < .05). RT6 generally performed good to excellent across activity intensities and monitor position (sedentary [AUC = 0.862–0.911], moderate [AUC = 0.849–0.830], vigorous [AUC = 0.872–0.877]) for non-dominant and dominant position, respectively. Cut-points were derived across activity intensities for non-dominant- and dominant-worn RT6 devices. Comparison of the RT6 derived cut-points identified appropriate agreement with comparative tools but yields the strongest agreement with the ActiGraph monitor at the hip location during sedentary, light, and moderate activity. Conclusion: The RT6 performed similar to the ActiGraph and GENEActiv and is capable of classifying the intensity of physical activity in young adults. As such this may offer a more useable tool for understanding current physical activity levels and in intervention studies to monitor and track changes without the excessive need for downloading and making complex analysis, especially given the option to view energy expenditure data while wearing it. The RT6 should be placed on the dominant hip when determining activities that are sedentary, moderate, or vigorous intensity.

Sign in / Sign up

Export Citation Format

Share Document