This study presents a comparative analysis of various curve-fitting methods for Cole’s Model, employed in electrical impedance spectroscopy (EIS) to characterize biological systems. The methods proposed by Ayllón (MA), Levenberg-Marquardt (LM), and González-Correa and Villanueva (M3P) were implemented and evaluated using a dataset of 159 bioimpedance measurement sets obtained from different devices. All methods were adapted and implemented in MATLAB to calculate the parameters of Cole’s Model and assess the normalized error be-tween experimental data and the generated fits. The results indicate that the methods specifically designed for Cole’s Model (MA, MAM, and M3P) demonstrate better performance in terms of normalized error compared to the LM method, which, although mathematically more robust, is not specifically tailored for this model. It is concluded that the computational implementation of these methods is effective and that, while the Levenberg-Marquardt method provides a good approximation, the specific methods yield superior results in error reduction.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Cole’s Model: A Comparative Study of Curve Fitting Methods

  • Natalia Yudit Bravo,
  • Agustín Domínguez,
  • Tomás Villanueva,
  • Guillermo Prisching,
  • Antonio H. Dell’Osa

摘要

This study presents a comparative analysis of various curve-fitting methods for Cole’s Model, employed in electrical impedance spectroscopy (EIS) to characterize biological systems. The methods proposed by Ayllón (MA), Levenberg-Marquardt (LM), and González-Correa and Villanueva (M3P) were implemented and evaluated using a dataset of 159 bioimpedance measurement sets obtained from different devices. All methods were adapted and implemented in MATLAB to calculate the parameters of Cole’s Model and assess the normalized error be-tween experimental data and the generated fits. The results indicate that the methods specifically designed for Cole’s Model (MA, MAM, and M3P) demonstrate better performance in terms of normalized error compared to the LM method, which, although mathematically more robust, is not specifically tailored for this model. It is concluded that the computational implementation of these methods is effective and that, while the Levenberg-Marquardt method provides a good approximation, the specific methods yield superior results in error reduction.