Methodology for preparing a cosmetic sample for the development of Microorganism Detection System (SDM) software and artificial intelligence learning to recognize specific microbial species.

ABSTRACT

INTRODUCTION AND OBJECTIVE: The article presents the methodology of preparing a cosmetic sample for analysi, and the creation of a dataset for teaching artificial intelligence to recognize specific species of microorganisms in cosmetic samples in terms of compliance with the ISO standard document, to develop of the Microorganism Detection System (SDM). MATERIAL AND METHODS: Methodology of preparation a cosmetic sample for testing covers the steps from taking a cosmetic sample to obtaining separated living microorganisms through staining to photos, which in the final stage are used for analysis of the purity of cosmetics by SDM, as well as for learning and testing of the deep convolutional neural network (CNN) for detecting and classifying cells of specific species of bacteria, fungi and yeast in cosmetics, according to the document of standard PN-EN ISO 17516-201411. RESULTS: A new techique was devised for preparing a cosmetic sample for the development of Microorganism Detection System (SDM) software, and artificial intelligence learning to recognize specific microbial species. Based on metod demonstrated, the Intelligent algorithms of SDM proved to be effective in counting and recognizing specific microorganisms (average accuracy for Candida albicans - 97%, Escherichia coli - 76%, Pseudomonas aeruginosa - 70%, Staphylococcus aureus - 85%), which are the most important species for the assessment of the purity of cosmetics. In addition, the reproducibility of the developed method was verified, and the results obtained were comparable to the breeding methods currently used, based on specific standards. CONCLUSIONS: The experiments confirmed the high sensitivity and specificity of the SDM method, its repeatability and, above all, the comparability of the results with clasic methods of European standards.