Research On The Predictive Model Of College Students' Cycling Energy Consumption Based On Actigraph GT3X

Objective:Based on the ActiGragh GT3X accelerometer,and based on the energy consumption measurements of K4b2 gas metabolism analyzer,in order to determine the best fitting position of the bicycle acceleration sensor,and establish the predictive equation of college students' bicycle energy consumption and MET,find the best critical point to differentiate VM axis acceleration counts for 4.8MET and 7.2MET.Methods:Selected 101 undergraduates and divided them into experimental group(81 persons)and validation group(20 persons)by sex and age.During the course of the experiment,subjects need wear K4b2 gas metabolic analyzer and GT3X acceleration sensor(Waist,thigh,ankle)at the same time;and take different intensity rides in the power of the bicycle(? lower strength:37%?65%VO2max;? medium strength:46%?63%VO2max;(? greater strength:64%?91%VO2 max).Through the use of descriptive statistics,correlation analysis and forced-entry regression analysis to compare the effectiveness of ActiGragh GT3X acceleration sensors in monitoring cycling energy expenditure at different wearing locations;then establish the prediction model of motion energy consumption by using stepwise regression method;And through the use of the correlation analysis,absolute error,relative error and the Bland-Altman point graph method to test the validity of the kinetic energy consumption prediction equation;At last,the best critical point of VM axis acceleration counting is established by ROC curve.Results:1)In each uniaxial,the average acceleration of the ankle position acceleration sensor ACx axis averaged the largest.The mean value of the VM axis counts of the ankle acceleration sensor is the largest among the 5 axes,and the most significant change of the induced acceleration.2)There was a significant correlation between the mean of counts of each axis of acceleration sensor and exercise energy consumption in three sites(P<0.01),among them the ACy axis of thigh(r = 0.83),the correlation coefficient of the ACz axis(r = 0.83)and the VM axis(r = 0.84)at the ankle was relatively high.3)In addition to the lumbar acceleration sensor Acx,ACh axis and thigh ACx axis,the other axes had a significant linear correlation with energy Met MET(P<0.01),the correlation coefficient between the ACz axis(r = 0.87),VM axis(r = 0.89)at the ankle and the MET is relatively high.4)The third-axis acceleration sensor at the ankle has the highest explanatory power of exercise(R2 = 0.70)and MET(R2 = 0.80).5)exercise energy consumption(kcal/min)=0.000219×VM+0.065×BW+0.145xSE-2.032(VM for the acceleration sensor coincidence counts value;BW for the weight(kg);SE for the gender(female = 0,male = 1));R2 is equal to 0.88,S and S/Y(%)are 0.61 and 11.55%,respectively.Through take the data of validation group test,we can get that the correlation coefficient between the predicted values and the measured values of K4b2 is between 0.82 and 0.86 at different levels of strength(P<0.01);The absolute error is 0.38?0.61 kcal/min,the relative error is 8.37%?10.54%,and 95%of the residuals fall within the range of Mean ? 1.96SD of the Bland-Altman scatter plot.6)Energy metabolism equivalent MET=0.00019007xVM+3.121,R2 is equal to 0.80,S is equal to 0.61,S/Y(%)is 10.33%.Through take the data of validation group test,we can get that the correlation coefficient between the predicted values and the measured values of K4b2 is between 0.80 and 0.85 at different levels of strength(P<0.01);The absolute error is 0.42?0.73,the relative error is 8.69%?9.61%,and 95%of the residuals fall within the range of Mean ± 1.96SD of the Bland-Altman scatter plot.7)The areas under 4.8MET and 7.2MET ROC curves were 0.926 and 0.901,respectively,and the sensitivities were 0.803,0.841 respectively.The VM axis acceleration counts have high diagnostic value for determining exercise intensity.The best threshold point of VM axis acceleration count corresponding to 4.8MET is 9764 counts/min,and the best critical point of VM axis acceleration count corresponding to 7.2MET is 21138 counts/min.Conclusion:1)The best fit for the ActiGragh GT3X accelerometer is at the ankle.2)Through wear the acceleration sensor in the ankle accumulate three variables of counts Value,Gender,Weight,established the energy consumption prediction can predict the energy consumption under different intensity levels effectively and can get higher prediction accuracy.3)The equation established by the VM axis counts of the acceleration sensor at the ankle can be used to effectively predict the MET of different energy intensity with higher prediction accuracy.4)The best cutoff points for the VM axis accelerations for 4.8MET and 7.2MET are 9764 counts/min and 21138 counts/min,respectively.