Quantitative Analysis of Fish Microbiological Quality Using Electronic Tongue Coupled with Nonlinear Pattern Recognition Algorithms

Abstract

Abstract The objective of this study was to establish quantitative evaluation models for fish microbiological quality analysis based on electronic tongue technique coupled with nonlinear pattern recognition algorithms. Crucian carp stored at 4 C were used. A commercial electronic tongue system was employed. The total viable counts ( TVC s) of fish samples were measured by the classical microbiological plating method. Partial least square regression, support vector regression ( SVR ) and back propagation neural network ( BP ‐ NN ) were applied comparatively to predict TVC values. The multivariate regression models were evaluated by the root mean square error of prediction ( RMSEP ) and the correlation coefficient in prediction set ( R pre ). Results revealed that the performance of BP ‐ NN model was superior to that of PLS model and SVR model. The RMSEP and R pre of the BP ‐ NN model for TVC prediction were 0.211 ln colony‐forming unit (cfu)/g and 0.993, respectively. This study showed that electronic tongue together with BP ‐ NN model could be a reliable technique for the detection of fish microbiological quality. Practical Applications Fish is a highly perishable commodity after harvesting and postmortem as a consequence of microbial breakdown mechanisms. Total viable count ( TVC ) method is the most widely used microbiological indicator for the evaluation of fish microbiological quality. However, the conventional analytical methods for the determination of TVC are cumbersome and time wasting. This work provides a practical and efficient way for rapid, accurate and convenient determination of TVC in fish using electronic tongue combined with regression algorithms to address these limitations.