Ranking factors affecting the decontamination efficacy of non-thermal plasma: The approach of dissipated power per plasma volume through machine learning modeling

Non-thermal plasma treatment can preserve food, but a meta-analysis assessing its efficacy has not been performed. This study retrieved the inactivation kinetics parameter log10D (n = 519), which varied largely among different plasma setups. Atmospheric pressure plasma jet, corona discharge, surface barrier discharge, dielectric barrier discharge, and inductively-coupled plasma had the highest efficacy with median D-values under 2 min. Dielectric barrier discharge, the most frequent setup (n = 160), was analyzed using dissipated power per plasma volume (W/cm3) as an integrated predictor of decontamination efficacy. Using conventional and machine learning approaches the most correlated parameters to the log10D were: dissipated power per plasma volume and matrix category, followed by microbial genus and pH. This study uses active learning to improve literature screening for data collection and various data analysis techniques for data treatment and ranking of the factors affecting non-thermal plasma decontamination. Non-thermal plasma decontamination shows potential but is also highly affected by the setup, and technical, biological, and food parameters. This study gives an overview of the decontamination performance that is to be expected for different matrix categories and setups, which could guide possible industrial applications. The ranking of the most important parameters to affect decontamination efficacy could be used for the optimization of these applications.