Jayaraman C, Mummidisetty C K, Jayaraman A
Annu Int Conf IEEE Eng Med Biol Soc. 2016 Aug;2016:3282-3285. doi: 10.1109/EMBC.2016.7591429.
Accuracy of physical activity estimates predicted by activity monitoring technologies may be affected by device location, analysis algorithms, type of technology (i.e. wearable/stickable) and population demographics (disability) being studied. Consequently, the main purpose of this investigation was to study such sensor dynamics (i.e. effect of device location, type and population demographics on energy expenditure estimates) of two commercial activity monitors. It was hypothesized that device location, population studied (disability), choice of proprietary algorithm and type of technology used will significantly impact the accuracy of the predicted physical activity metrics. 10 healthy controls and eight individuals with spinal cord injury (SCI) performed structured activities in a laboratory environment. All participants wore, (i) three ActiGraph-G3TX's one each on their wrist, waist & ankle, (ii) a stickable activity monitor (Metria-IH1) on their upper-arm and (3) a Cosmed-K4B metabolic unit, while performing sedentary (lying), low intensity (walk 50 steps at self-speed) and vigorous activity (a 6 minute walk test). To validate the hypothesis, the energy expenditures (EE) predicted by ActiGraph-GT3X and Metria-IH1 were benchmarked with estimated EE per Cosmed K4B metabolic unit. To verify the step count accuracy predicted by ActiGraph-GT3X's and Metria-IH1, the manually calculated step count during the low intensity activity were compared to estimates from both devices. Results suggest that Metria-IH1 out-performed ActiGraph-GT3X in estimating EE during sedentary activity in both groups. The device location and population demographics, significantly affected the accuracy of predicted estimates. In conclusion, selecting activity monitor locations, analysis algorithm and choice of technology plays based on the movement threshold of population being studied can pave a better way for reliable healthcare decisions and data analytics in population with SCI.
活动监测技术所预测的身体活动估计值的准确性，可能会受到设备位置、分析算法、技术类型（即可穿戴式/粘贴式）以及所研究人群的人口统计学特征（残疾情况）的影响。因此，本研究的主要目的是研究两款商用活动监测器的此类传感器动态特性（即设备位置、类型和人群人口统计学特征对能量消耗估计值的影响）。研究假设为，设备位置、所研究的人群（残疾情况）、专有算法的选择以及所使用的技术类型，将显著影响预测的身体活动指标的准确性。10名健康对照者和8名脊髓损伤（SCI）患者在实验室环境中进行了结构化活动。所有参与者在进行久坐（躺卧）、低强度（以自身速度行走50步）和剧烈活动（6分钟步行测试）时，均佩戴了：(i) 三个ActiGraph-G3TX，分别戴在手腕、腰部和脚踝；(ii) 一个粘贴式活动监测器（Metria-IH1）在上臂；以及(3) 一个Cosmed-K4B代谢单元。为了验证该假设，将ActiGraph-GT3X和Metria-IH1预测的能量消耗（EE）与每个Cosmed K4B代谢单元估计的EE进行了对比。为了验证ActiGraph-GT3X和Metria-IH1预测的步数准确性，将低强度活动期间手动计算的步数与两款设备的估计值进行了比较。结果表明，在两组的久坐活动中，Metria-IH1在估计EE方面优于ActiGraph-GT3X。设备位置和人群人口统计学特征显著影响了预测估计值的准确性。总之，根据所研究人群的运动阈值选择活动监测器的位置、分析算法和技术类型，可为脊髓损伤人群的可靠医疗决策和数据分析铺平更好道路。
