Artificial neural networks (ANN) are non-linear calibration methods that try to simulate the human nervous system with neurons, layers and transfer functions. In this study the input neurons were the UV-VIS-NIR spectra of egg yolks and the outputs were cholesterol content. Therefore, the aim of this work was to develop and assess a UV-VIS-NIR method coupled to chemometrics and ANN for cholesterol determination in egg yolks. A total of 57 samples of egg yolks were obtained from fresh shell eggs and pasteurised egg yolks. A total of 2311 variables (wavelengths) with four variable selection methods were analysed. An enzymatic method was used as reference method for cholesterol content. The effect of the solvent was also evaluated. The best results were obtained based on selected absorbance peaks. Calibration results showed r2cal over 0.90 and RMSEC below 0.65. Validation results showed r2val close to 0.7. The ANN model developed could be useful to determine cholesterol in egg yolk for routine quality control.
Read full abstract