Table 4 Technical Data for analysis phase in Typically Developing Children
Author | Accelerometer data comparison | Differences between the two hands | Data cleaning | Threshold (cutoff frequency of filter applied on raw data) | Threshold to assess the intensity of arm movement |
|---|---|---|---|---|---|
[25] Birmingham A. T. et al. (1985) | RMS of tremor amplitude, dominant peak and its frequency. | For rest tremor, amplitude in the dominant hand was significantly lower in adolescence and early adult life than in childhood, for the non-dominant hand the statistically significant difference was sustained to later life. For work tremor, dominant hand frequency declined significantly with age, both hands continue to decline in adulthood. | Frequency analysis of the tremor waveform was filtered to remove frequencies above 50Â Hz to prevent alias contamination | 50Â Hz | NA |
[24] Avi Sadeh et al. (1994) [Study 1] | Accelerometric data matched with PSG scoring performed to develop the scoring algorithm: PS probability of sleep | The mean activity level of the dominant wrist was significantly higher than that of the nondominant wrist during PSG-determined sleep (6.84 vs. 6.16), as well as during wakefulness (25.8 vs. 22.3). | NA | NA | NA |
[27] Deutsch K. M. et al. (2006) | The peak frequency within two frequency bands (5–15 Hz and 15–30 Hz) and the proportion of power exhibited at the peak frequency determined (based on power spectral density calculated using Welch’s averaged periodogram method). | The peak frequency of the finger of the dominant hand (21.4 Hz) was higher than nondominant hand (20.7 Hz) in the 15–30 Hz frequency band. No significant differences in proportion of power exhibited at peak frequency within the 5–15 Hz of postural tremor as a function of age, hand dominance or hand configuration. Postural tremor of nondominant hand was significantly more regular than dominant hand. | Band-pass filtered | 1 Hz - 50 Hz | NA |
[34] Graves L.E.S. et al. (2008) | Means and standard deviations of activity counts (counts/min) | Activity of the dominant limb was significantly greater than non-dominant during tennis and bowling (P < 0.001) and non-dominant limb activity was significantly greater during boxing than bowling or tennis (P < 0.001). Activity counts from the left wrist for tennis and boxing (r = 0.710 and 0.744, P < 0.01) and the right wrist for boxing (r = 0.586, P < 0.05) were significantly correlated with EE. | Band pass filtering | 0.21–2.28 Hz | NA |
[17] Davila E. M. (2011) | Data Trasformation: AEE, Time. Data Summarization Characteristics: Bouts Duration, Intensity Thresholds. | No statistical differences between outcome variables for any bout duration (1, 5, 10 min) within L and MV intensity categories between AMs (D versus ND, LW versus RW) or model (1R versus 2R). Dominant and RW AMs were no-significantly higher than ND and LW, respectively, within MVPA intensity. In contrast, ND and LW AMs were non-significantly higher than D and RW within L intensity PA. Identical results within gender. | Quantity control checks were performed to identify periods on non-wear. | NA | Light (AEE < 0.05 kcals/kg/min), moderate (0.05 < AEE < 0.09 kcals/kg/min), vigorous (AEE ≥ 0.10 kcals/kg/min). |
[21] Phillips L. R. S. et al. (2012) | VM with gravity-substracted. | Both sides demonstrated good criterion validity (right: r = 0.9, left: r = 0.91) and good concurrent validity (right: r = 0.83, left: r = 0.845). ROC analysis proved GENEA monitors able to successfully discriminate among all intensity levels. | NA | NA | Sedentary (<  1.5 METs), light (1.5–2.99 METs), moderate (3–5.99 METs) and vigorous (≥ 6 METs). The accelerometer counts for activities were coded into binary indicator variables (0 or 1) based on intensity. |
[28] MacArthur B. et al. (2014) | Percentage of time in MVPA calculated by summing the number of 15-s intervals in which the activity counts were ≥ 574 counts/15 s. | The accelerometers placed on the wrists did not find differences in the conditions in percentage MVPA (right: 48.8 ± 29.5%, left: 47.6 ± 28.8%). | NA | NA | MVPA: activity counts ≥574 counts/15 s. |
[19] Lemmens R. J. M. et al. (2015) | ICC parameter (based on VM). | Within-subject reliability calculated for the 2 arm hands separately, median ICCs ranged between 0.68–0.92. Between subject reliability for the 2 arm hands separately, median ICCs ranged between 0.61–0.90. | Zero time-phase, low-pass filtered | 1.28 Hz | NA |
[31] Kaneko M. et al. (2015) | Postural stability of the hands and elbows, rotational speed, mirror movement, two parameters of bimanual symmetry, compliance | All indices had a tendency to increase with age. | Low-pass filter | 6Â Hz | NA |
[35] Dadashi F. et al. (2016) [Group 2] | Average propulsive phases of right and left arms, pull and push phases (Δpull, Δpush), sum of aerial recovery and entry catch phases (ΔNProp), index of coordination (IdC). | By increasing the velocity, the duration of arm under-water phases (Δpull + Δpush) and accordingly IdC did not change significantly. G2 group used 2,8% lower catch-up pattern (P < 0,01) by increasing the arm under-water phases (P < 0.016) and using 6.5 more arm stroke (P < 0.001). No changes in the stroke length and cycle velocity variation were observed (P > 0.22). | NA | NA | NA |
[36] Mackintosh K.A. et al. (2016) | Mean and variance of the accelerometer counts in each 15 s window. These extracted features were used as inputs into the ANNs, a specific type of machine learning model. RMSE. | The ANNs for left and right wrist accelerometers had a lower correlations with predicted EE. No significant differences in RMSE analysis. Despite significant advantages in terms of compliance, they could lead to potentially marginal losses in EE prediction accuracy. | NA | NA | 1,4% of collected data were removed when EE < 0,5 MET (measured with MetaMax 3B) |