Neural network model for growth of Salmonella serotypes in ground chicken subjected to temperature abuse during cold storage for application in HACCP and risk assessment

Authors: Oscar, Thomas

Submitted to: International Journal of Food Science and Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/5/2016
Publication Date: 1/1/2017
Citation: Oscar, T.P. 2017. Neural network model for growth of Salmonella serotypes in ground chicken subjected to temperature abuse during cold storage for application in HACCP and risk assessment. International Journal of Food Science and Technology. 52:214-221. doi: 10.1111/ijfs.13242.

Interpretive Summary: Models that predict growth of foodborne pathogens in unit operations (e.g. cold storage) of the food production chain are valuable tools for food safety. With the advent of commercial software applications, it is now easy to develop models for predictive microbiology applications. We characterized the behavior of Salmonella in ground chicken thigh meat stored at 61F and found that the growth of this deadly pathogen differs among eight subtypes of this bacteria. Consequently, commercial software applications were used to develop three versions of a model for growth of Salmonella in ground chicken thigh meat. All versions of the model will be deployed on the Poultry FARM website (www.ars.usda.gov/naa/errc/PoultryFARM) for use by the food industry, risk assessors, regulators, other scientists, and consumers at no cost to these customers.

Technical Abstract: With the advent of commercial software applications, it is now easy to develop neural network models for predictive microbiology applications. However, different versions of the model may be required to meet the divergent needs of model users. In the current study, the commercial software applications Excel, NeuralTools, and @Risk were used to develop two deterministic and one stochastic version of a neural network model for growth of Salmonella serotypes in ground chicken thigh meat stored for 0 to 8 days at 16 degrees C. An automated miniature most probable number (MPN) method was used to enumerate Salmonella serotypes (n = 8) in ground chicken thigh meat portions during storage for 0, 1, 2, 4, 6, or 8 days at 16 degrees C. Growth of Salmonella was affected (P < 0.05) by serotype at 4, 6, and 8 days of storage but not at 0, 1, or 2 days of storage at 16 degrees C. When results for 4, 6, and 8 days of storage at 16 degrees C were combined, the log MPN per portion ranged from 6.12 plus or minus 0.47 (mean plus or minus SD) for serotype 8,20:-:z6 to 6.84 plus or minus 0.23 for serotype Thompson. A deterministic, multiple-layer feedforward neural network model with a single hidden layer of two nodes was developed using Excel and NeuralTools. Performance of the model was evaluated using the acceptable prediction zone method. The proportion of residuals (observed – predicted) in an acceptable prediction zone (pAPZ) from -1 log (fail-safe) to 0.5 log (fail-dangerous) was 0.948 for training data (n = 192) and 0.988 for testing data (n = 84. A pAPZ less than or equal to 0.7 indicated that the model provided predictions with acceptable bias and accuracy. Thus, the model was successfully validated. Two additional versions of the model were developed. One was a deterministic version that did not require NeuralTools to run and another was a stochastic version for use in risk assessment that required NeuralTools and @Risk to run.