Thermal inactivation of Salmonella in ready-to-use therapeutic food and prediction using neural network and conventional models

Ready-to-use therapeutic food (RUTF) is a low-moisture, lipid-rich product widely used for the treatment of severe acute malnutrition worldwide, which calls for heat processing to ensure its microbiological safety. This study investigated the thermal inactivation of Salmonella in a pre-cooked RUTF product and developed neural network (NN) models to predict Salmonella die-off. Commercial pre-cooked RUTF was inoculated with a cocktail of three Salmonella serotypes that were associated with previous nut outbreaks and recalls in the U.S. Inoculated RUTF (water activity = 0.27 ± 0.002) was heat-treated for up to 90 min at 75–95 °C, which was selected as a practical heat treatment range for evaluation in this study, informed by industry inputs from commercial RUTF manufacturers worldwide. Experimental results indicated that 2.7 log CFU/g of Salmonella was reduced at 75 °C after 90 min (D 75°C = 33.4 min), while >5 log reduction was achieved at 85 °C after 60 min (D 85°C = 10.3 min) and > 7 log reduction was obtained at 95 °C after 30 min (D 95°C = 4.1 min). The NN models demonstrated accurate prediction of Salmonella reduction during heat treatments, with a small root mean square error (RMSE) of 0.23 log CFU/g and a high R 2 of 0.989. Model validation further confirmed model accuracies (RMSE = 0.36 log CFU/g and R 2 = 0.986). This study provides heat treatment data and procedures for RUTF products and offers a potential NN model for Salmonella inactivation, which could serve as key scientific evidence to establish key performance indicators in RUTF formulation and manufacturing.