This paper addresses the nonlinear inverse problem of estimating the parameters of the Cole–Cole model used to describe the behavior of the complex permittivity of blood samples. Such a model provides an efficient and accurate representation of biological tissues in the entire frequency band considered and reduces the complexity of the experimental data to a few parameters. In this way, it is possible to extract a “synthetic view” of the dielectric properties of tissues in such a way that more information on the glucose concentration can be derived, in addition to the resonance peak or phase shift. In order to perform the fitting of the Cole–Cole model, two different algorithms were used and compared: the Levenberg–Marquardt and the variable projection algorithms. Synthetic data present in the literature were used to evaluate the performances obtainable with these methods. In particular, Monte Carlo analysis was used in order to evaluate the accuracy and the precision that these two methods provide in the process of estimating the parameters involved, with respect to the starting points of the parameters. The results obtained showed that the variable projection algorithm always outperformed the Levenberg–Marquardt one, although the former has a greater computational burden than the latter.

Comparing two Fitting Algorithms for Determining the Cole–Cole Parameters in Blood Glucose Problems

Dima, Roberto;Buonanno, Giovanni;Solimene, Raffaele
2021

Abstract

This paper addresses the nonlinear inverse problem of estimating the parameters of the Cole–Cole model used to describe the behavior of the complex permittivity of blood samples. Such a model provides an efficient and accurate representation of biological tissues in the entire frequency band considered and reduces the complexity of the experimental data to a few parameters. In this way, it is possible to extract a “synthetic view” of the dielectric properties of tissues in such a way that more information on the glucose concentration can be derived, in addition to the resonance peak or phase shift. In order to perform the fitting of the Cole–Cole model, two different algorithms were used and compared: the Levenberg–Marquardt and the variable projection algorithms. Synthetic data present in the literature were used to evaluate the performances obtainable with these methods. In particular, Monte Carlo analysis was used in order to evaluate the accuracy and the precision that these two methods provide in the process of estimating the parameters involved, with respect to the starting points of the parameters. The results obtained showed that the variable projection algorithm always outperformed the Levenberg–Marquardt one, although the former has a greater computational burden than the latter.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11591/615684
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact