Methods to improve the accuracy of spectrophotometer determination of serum creatinine

Abstract

In order to solve a series of problems that affect the accuracy of spectral quantitative analysis of complex solution components, based on the “M + N” theory, this paper proposes a system for quantitative analysis of serum components by spectroscopy. At the same time, feasible methods are used in each step of the system to effectively reduce the influence of noise and interference factors, and obtain the spectral signal of serum components with high signal-to-noise ratio, so that the model has good prediction performance. In this paper, serum creatinine was taken as the research object. This paper collects multi-location and multi-mode spectral data, and uses the wavelength elimination method to optimize the wavelength to avoid over-fitting, then, based on PLS modeling, the cubic fitting is used to optimize the model, so as to reduce the influence of spectral nonlinearity caused by the scattering characteristics of solution components. After using the above methods, the prediction ability and robustness of the model are improved. Compared with the results without these methods, the correlation coefficient of creatinine concentration of all samples predicted by the model established by using the reasonable methods is increased by 32.43%, and the root mean square error is reduced by 53.61%. The above experiments show that this article combines multi-mode spectrum, the wavelength variable optimization method, the cubic equation optimization model method and other methods to quantitatively analyze the complex solution components based on the spectrum, which can improve the accuracy and robustness of the results.