Linear Regression Models for Fat Free Mass Measurement Based on Bioelectrical Impedance Analysis

Abstract

This work proposes new mathematical models for fat free mass (FFM) measurements for both male and female people based on bioelectrical impedance analysis. In this research total 2650 (1300 male and 1350 female) data have been used. Bioelectrical impedance at 50 kHz along with physical parameters like age, height, body mass index (BMI) have been used for the mathematical model development. The proposed models have been analyzed statistically by evaluating correlation (Pearson) coefficients, 95% limit of agreement (LOA), absolute errors, bias, root mean square error (RMSE) and compared with actual data. The results show that the correlation (Pearson) coefficients are $0.998 (\mathrm{p}\lt 0.001)$ for male and $0.997 (\mathrm{p}\lt 0.001)$ for female people which indicate very good matching with actual data. The intervals of LOA are only -1.93 kg to 1.80 kg and -1.48 kg to 1.38 kg for male and female data respectively and most of errors found by proposed models remain within the limit of agreement. The absolute errors (mean ± Standard Deviation) are $(0.81 \pm 0.50)$ kg and $(0.62 \pm 0.38)$ kg for male and female data respectively whereas the bias are -0.06 kg for male and -0.05 kg for female population only. The RMSE are also very low and which are 0.95 kg for male and 0.73 kg for female people. Comparing the results of this research with existing models it is seen that proposed mathematical models exhibit better results and hence the proposed models can be more suitable for FFM measurement.