Application of artificial neural network to predict Escherichia coli O157:H7 inactivation on beef surfaces. (January 2015)

• Record Type: Journal Article
• Authors: Gosukonda, Ramana; Mahapatra, Ajit K.; Liu, Xuanli; Kannan, Govind
• Is Part Of: Food control. Volume 47(2015:Jan.)
• Page Start: 606
• Page End: 614
• Publication Date: 2015-01
• Subjects: Artificial neural network -- Predictive modeling -- E. coli O157:H7 -- Low-voltage direct electric current -- Beef -- Inactivation; Food -- Quality -- Periodicals; Food -- Analysis -- Periodicals; Food handling -- Periodicals; Food industry and trade -- Quality control -- Periodicals; Aliments -- Industrie et commerce -- Qualité -- Contrôle -- Périodiques; Aliments -- Qualité -- Périodiques; Aliments -- Analyse -- Périodiques; Hygiène alimentaire -- Périodiques; Food -- Analysis; Food handling; Food -- Quality; Periodicals; Electronic journals; 664.07
• DOI: 10.1016/j.foodcont.2014.08.002
• Languages: English
• ISSNs: 0956-7135

Abstract

Abstract: The objective of this study was to develop artificial neural network (ANN) models for quantifying Escherichia coli O157:H7 ( E. coli ) inactivation due to low-voltage electric current on beef surfaces and to compare them with statistical models for their suitability as a tool for online processing by the meat industry. Modeling techniques with optimal prediction accuracies of E. coli inactivation on meat would not only enhance the meat quality and public perception from a safety perspective, but also improve the marketability of the meat products. The data used in this study were obtained from experiments that measured the percentage (%) of E. coli O157:H7 reduction (output) on beef surfaces when subjected to current (input 1) 300, 600, and 900 mA, duty cycles (input 2) 30, 50, and 70%, and frequency (input 3) 1, 10, and 100 kHz for three treatment times (2, 8, 16 min). Data were subjected to statistical and artificial neural network (ANN) modeling techniques. Data from each input set were sub-partitioned into training, testing, and validation data sets for ANN. Back-propagation (BP) and Kalman filter (KF) learning algorithms were used in ANN to develop nonparametric models between input and output data sets. The trained ANN models were cross-tested with validation data. Various statistical indices including R 2 between actual and predicted outputs were produced and examined for selecting the best networks. Prediction plots for current, frequencies, and duty cycles indicated that ANN models had better accuracies compared to the statistical models in predicting from unseen pattern. Further, ANN models were able to more robustly generalize and interpolate unseen patterns within the domain of training. Since ANN models have the inherent ability to handle high biological variability and the uncertainty associated with inactivation of microorganisms, they have great potential for meat quality evaluation and monitoring in meat industry. Highlights: ANN models were developed for predicting E. coli inactivation. ANN models were compared with statistical models. ANN models had a better accuracies compared to the statistical models. ANN models were able to more robustly generalize and interpolate unseen patterns. ANN models would enable to improve meat quality in the meat industry.