Open Access

Validation of non-empirical fat-free mass estimation model for a wrist-worn device

Aleksandr Polokhin

Aleksandr Polokhin

AURA Devices, IncWilmington, USA
Search for this author on
Sciendo|Google Scholar
Polokhin, Aleksandr
, 
Anna Pronina

Anna Pronina

AURA Devices, IncWilmington, USA
Search for this author on
Sciendo|Google Scholar
Pronina, Anna
, 
Andrey Boev

Andrey Boev

AURA Devices, IncWilmington, USA
Search for this author on
Sciendo|Google Scholar
Boev, Andrey
 and 
Stas Gorbunov

Stas Gorbunov

AURA Devices, IncWilmington, USA
Search for this author on
Sciendo|Google Scholar
Gorbunov, Stas
  
Jun 25, 2022
Validation of non-empirical fat-free mass estimation model for a wrist-worn device's Cover Image
About this article
Previous Article
Next Article
Cite
Download Cover
Published Online: Jun 25, 2022
Page range: 31 - 38
Received: Apr 08, 2022
DOI: https://doi.org/10.2478/joeb-2022-0006
Keywords
© 2022 Aleksandr Polokhin, Anna Pronina, Andrey Boev, Stas Gorbunov, published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.

Fig.1

Flow chart of the study participants included in the analysis.
Flow chart of the study participants included in the analysis.

Fig.2

The positions of the WIM electrodes on the left wrist and right hand during bioelectrical impedance measurement.
The positions of the WIM electrodes on the left wrist and right hand during bioelectrical impedance measurement.

Fig.3

The hand-to-hand bioelectrical impedance scheme of the TTC model.
The hand-to-hand bioelectrical impedance scheme of the TTC model.

Fig.4

Arm volume multiplied by TBM/FFM ratio approximates bioimpedance: (a) Approximation of InBody 230 arm impedance. (b) Approximation of real part of bioimpedance measured with WIM.
Arm volume multiplied by TBM/FFM ratio approximates bioimpedance: (a) Approximation of InBody 230 arm impedance. (b) Approximation of real part of bioimpedance measured with WIM.

Fig.5

(a) MAPE; (b) Coefficient of determination r2; (c) Maximal residual error. Distribution shapes were obtained during 100,000-sample bootstrap procedure. Bold lines show normal approximation curves; (d) The effect of sex and age variables on maximal residual error distribution for the single cylinder model.
(a) MAPE; (b) Coefficient of determination r2; (c) Maximal residual error. Distribution shapes were obtained during 100,000-sample bootstrap procedure. Bold lines show normal approximation curves; (d) The effect of sex and age variables on maximal residual error distribution for the single cylinder model.

Fig.6

Bland-Altman plots for WIM FFM estimate validation: (a) Single cylinder (SC) model: confidence intervals for the mean bias: (-0.874, 0.879) kg, for upper LOA: (4.380, 6.937) kg, for lower LOA: (-6.932, -4.375) kg; TTC model: for mean bias: (-0.920, 0.905) kg, for upper LOA: (4.547, 7.208) kg, for lower LOA: (-7.223, -4.562) kg. (b) Skinfold thickness calculations: confidence intervals for the mean bias: (-1.792, 0.955) kg, for upper LOA: (6.653, 10.651) kg, for lower LOA: (-11.488, -7.4918) kg.
Bland-Altman plots for WIM FFM estimate validation: (a) Single cylinder (SC) model: confidence intervals for the mean bias: (-0.874, 0.879) kg, for upper LOA: (4.380, 6.937) kg, for lower LOA: (-6.932, -4.375) kg; TTC model: for mean bias: (-0.920, 0.905) kg, for upper LOA: (4.547, 7.208) kg, for lower LOA: (-7.223, -4.562) kg. (b) Skinfold thickness calculations: confidence intervals for the mean bias: (-1.792, 0.955) kg, for upper LOA: (6.653, 10.651) kg, for lower LOA: (-11.488, -7.4918) kg.

Anthropometric, body composition and bioelectrical impedance data_

Mean ± STD Range
Age, years 27.02 ± 10.2 18 - 56
TBM, kg 69.56 ± 11.29 45.83 - 100.95
Height, cm 171.89 ± 8.86 156.0 - 196.0
BMI, kg/m2 23.6 ± 3.97 18.17 - 38.87
Total arms 6109.06 ± 1505.04 3480.6 - 10970.2
volume, cm3
FFMDEXA, kg 50.99 ± 10.94 34.51 - 86.59
FMDEXA, % 26.77 ± 9.34 8.32 - 48.39
Re(Z)10_a, Ohm 633.13 ± 136.83 366.43 - 1001.8
Re(Z)10_b, Ohm 559.43 ± 169.33 88.28 - 866.89
Re(Z)10_c, Ohm 627.71 ± 133.16 398.16 - 964.93
Re(Z)50_a, Ohm 592.86 ± 134.54 379.79 - 914.78
Re(Z)50_b, Ohm 551.97 ± 160.0 126.64 - 889.12
Re(Z)50_c, Ohm 562.21 ± 138.21 304.83 - 831.84
Re(Z)100_a, Ohm 571.13 ± 133.97 352.69 - 881.58
Re(Z)100_b, Ohm 553.53 ± 141.96 312.6 - 890.5
Re(Z)100_c, Ohm 527.9 ± 134.54 204.77 - 799.5

Linear regression metrics for FFM estimation: two models (SC and TTC) and nine experimental conditions

Frequency, kHz Right hand electrode position SС model
 TTC model
r2 MAPE Max. err. r2 MAPE Max. err.
10 a 0.656 ± 0.209 0.076 ± 0.016 15.732 ± 5.076 0.679 ± 0.211 0.075 ± 0.017 14.404 ± 4.756
10 b 0.605 ± 0.487 0.085 ± 0.023 15.088 ± 6.327 0.514 ± 0.37 0.096 ± 0.022 17.862 ± 6.684
10 c 0.777 ± 0.114 0.065 ± 0.012 12.463 ± 4.224 0.795 ± 0.112 0.064 ± 0.013 11.236 ± 3.888
50 a 0.763 ± 0.119 0.064 ± 0.014 12.493 ± 3.357 0.774 ± 0.125 0.062 ± 0.014 12.16 ± 3.456
50 b 0.479 ± 0.309 0.111 ± 0.309 27.447 ± 24.079 0.422 ± 0.338 0.119 ± 0.338 30.852 ± 25.537
50 c 0.876 ± 0.054 0.051 ± 0.009 8.623 ± 2.596 0.886 ± 0.051 0.052 ± 0.009 6.884 ± 1.283
100 a 0.752 ± 0.131 0.065 ± 0.014 12.883 ± 3.534 0.762 ± 0.14 0.064 ± 0.015 12.595 ± 3.507
100 b 0.698 ± 0.143 0.078 ± 0.015 13.653 ± 3.794 0.761 ± 0.133 0.07 ± 0.014 11.829 ± 2.888
100 c 0.482 ± 0.423 0.086 ± 0.02 20.076 ± 9.376 0.435 ± 0.461 0.094 ± 0.021 20.44 ± 10.855
Language:
English
Publication timeframe:
1 times per year
Journal Subjects:
Engineering, Bioengineering and Biomedical Engineering, Biomedical Electronics, Life Sciences, Biophysics, Medicine, Biomedical Engineering, Physics, Spectroscopy and Metrology
Journal RSS Feed