Research On Novel Measuring And Modeling Methods For Electrical Impedance Myography

Electrical impedance myography(EIM),proposed by Seward B.Rutkove,a professor in Harvard Medical School,is a new method for the diagnosis of neuromuscular disease based on measuring and modeling analysis of muscular electrical impedance.It has the advantages for non-radiative,non-invasive,simple and low-cost measurements,and therefore could have a potential application prospect in the diagnosis of neuromuscular disease,including early detection,long-time monitoring,treatment and drug development,etc.However,the existing EIM methods cannot simultaneously meet the requirement for high-speed and accurate measurement.First,the single-frequency(SF)EIM analysis only includes the characteristics of 50 k Hz muscular electrical impedance,which cannot be used to accurately model and analyze the characteristics of muscular broadband impedance spectroscopy.Second,the frequency-sweep EIM analysis can be used to estimate more characteristics of muscular impedance in a wide frequency range,from which the corresponding equivalent circuit model can be established.However,frequency-sweep is usually timeconsuming and could induce medium polarization in the measurement process.Moreover,the estimation of equivalent model parameters usually requires iterative optimization,in which the improper selection of the iterative initial values could easily lead the optimization algorithm to premature convergence,thus affecting the speed and accuracy on EIM modeling.Therefore,how to further improve measurement technology,optimize modeling theory,and establish a fast and accurate method for EIM analysis is of great significance to its extension on the diagnosis of neuromuscular disease.To solve the problems mentioned above,the main research content of the dissertation includes the following parts:(1)Research on novel SF-EIM methods based on complete-response measurement(transient response and steady-state response)for integer-order and fractional-order EIM modeling.Here,both numerical analysis and curve-fitting methods in the time-domain are proposed for fast and accurate estimate of the integer-order Fricke-Morse model using a 50 k Hz sinusoidal excitation,and the fractional-order Cole model using a 50 k Hz DC-biased sinusoidal excitation.First,the mathematical expression of the complete-response voltage is deduced from the inverse Laplace transform of the product of the(DC-biased)sinusoidal current excitation and the corresponding EIM model.Second,the characteristics of the complete-response signal in the time-domain and frequency-domain are estimated based on transient and steady-state analysis,from which the analytical expressions of the model parameters are obtained.Finally,the sum-of-squared-error between the measured completeresponse voltage and the calculated model-based voltage is minimized through an iterative optimization algorithm,in which the model parameters estimated by the previous analytical method are used as the iterative initial values,in order to reduce the influence on the speed and accuracy of the iterative optimization process;(2)Research on novel multi-frequency(MF)EIM methods based on integer-period steady-state response measurement for integer-order and fractional-order EIM modeling.Here,a MF-EIM method based on curve-fitting in the frequency-domain using a multisine excitation and its steady-state response on integer-period synchronous measurement is proposed to fast and accurate estimate of the feature parameters of the aforementioned integerorder and fractional-order equivalent impedance models.First,the rationale for integerperiod synchronous impedance spectroscopy measurement using a multisine excitation is explained,which could be used to fundamentally eliminate the effect of spectrum leakage in broadband EIM analysis.Second,an optimization algorithm for crest-factor minimization of a multisine with sparse spectrum distribution is proposed,from which a near-binary multisine(NBM)is synthesized as the excitation signal for improving the SNR of integerperiod synchronous broadband EIM measurement.Finally,a fast curve-fitting method is proposed to estimate the model parameters using matrix operation-based least square algorithm,in which the curve parameters of the equivalent impedance model on the Nyquist plot can be fitted to minimize the sum-of-squared-error between the measured impedance spectroscopy and the data calculated from the corresponding model.Therefore,the matrix operation-based fitting algorithm could overcome some drawbacks existing in the former iterative fitting algorithm,e.g.,premature convergence and time-consuming;(3)Simulation and experimental analysis for the novel EIM measuring and modeling methods.Here,different simulations are conducted first,and then a high-speed EIM measurement system based on a novel broadband mirrored current source is established to verify the effectiveness of the above-mentioned SF/MF EIM measuring and modeling methods.Several three-element(two resistors and one capacitor,2R-1C)circuits and typical in-vivo muscles used in EIM analysis are chosen as the test samples for the SF-based complete response measurement,MF-based electrical impedance spectroscopy measurement,and feature recognition of the corresponding equivalent integer-order and fractional-order circuit models.Simulation results show that the relative error of the estimated model parameters are all less than 1% in both the proposed SF/MF-EIM analysis methods and the traditional frequency-sweep method under 40 d B Gaussian white noise,but the measurement time for SF and MF analysis is only 0.295 ms and 1 ms,respectively,which is far less than 19 ms,the measurement time for the frequency-sweep method.In addition,the relative error of the estimated model parameters from the new instrument using SF measurement is less than 5%on the analysis of standard 2R-1C circuit,which is slightly larger than 2%,the result obtained from both the new instrument using MF measurement and the a standard instrument(SFB7)using frequency-sweep measurement.Besides,the estimated results show that the Cole model(average RMSE less than 2%)could be more accurate than the Fricke-Morse model(average RMSE less than 12%)in modeling the muscle tissue.The RMSE of the estimated Fricke-Morse model,however,could be lower than 3% using single-frequency analysis compared to broadband or frequency-sweep analysis.In sum,the methods presented could be beneficial to develop and improve the technology and theory for EIM analysis,and have broad application prospect especially in simplifying the measurement circuit,reducing the instrument cost,increasing the accuracy and the speed,etc.It could provide new ideas for the development of novel EIM analyzer and for the subsequent applications on the diagnosis of neuromuscular disease.