Preliminary study on parameterization of raw electrical bioimpedance data with 3 frequencies.

This study tests the geometrical parameterization method for Electrical Bio-Impedance Spectroscopy (EBIS) readings previously proposed by one of the authors. This method uses the data of just three frequencies (therefore called 3P method). The test was carried out by the analysis of parameterization from 26 spectra (selected from 13 data sets) by the non-linear square (NLS) method, the 3P method and a combination of the two (3P-NLS). Additionally, the behaviour of the 3P method for 4 levels of noise and 3 different ways of segmenting the spectra were also explored with a MATLAB simulation of 400 spectra. Finally, a system for the classification of EBIS readings is presented, based on deviations of the raw data from the semi-circle obtained by the parameterization methods. Overall, the results suggest a very good performance of the 3P method when compared with the other two. The 3P method performs very well with levels of noise of 1 and 2%, but performs poorly with levels of noise of 5% and 10%. The results support the idea that the 3P method could be used with confidence for the parameterization of EBIS spectra, after the selection of three adequate frequencies according to specific applications.