Artificial intelligence and machine learning applications in food microbiology: A systematic review

Abstract

The integration of artificial intelligence (AI) and machine learning (ML) into food microbiology offers transformative potential for enhancing food safety through rapid pathogen detection. Traditional microbiological methods, while reliable, are time-consuming and ill-suited for analyzing complex, high-dimensional datasets. This systematic review synthesizes contemporary evidence on AI/ML applications in food microbiology, specifically evaluating their performance in microbial detection, contamination prediction, and spoilage assessment, while identifying barriers to real-world implementation. Following PRISMA 2020 guidelines, we systematically searched PubMed, Scopus, Web of Science, and IEEE Xplore for studies published between January 2017 and December 2023. Data were extracted on model types, performance metrics (accuracy, F1-score, AUC-ROC), and validation approaches. From 1,153 records screened, 22 studies met the inclusion criteria. Supervised learning dominated (91% of studies), with Random Forest (n=9), Convolutional Neural Networks (CNNs; n=6), and Support Vector Machines (n=5) being most prevalent. Applications focused on: (1) rapid pathogen detection from hyperspectral/multispectral imaging (accuracy: 89-96%; AUC: 0.88-0.95); (2) prediction of microbial growth kinetics under varying storage conditions (RMSE: 0.15-0.45 log CFU/g); and (3) spoilage classification from volatile organic compound patterns (F1-score: 0.85-0.93). AI/ML models demonstrate strong analytical performance in controlled settings but face significant translation challenges, including data scarcity, model interpretability, and integration with existing workflows. Future research should prioritize standardized benchmark datasets, explainable AI approaches, and validation in operational food industry environments to bridge the lab-to-field gap.