Colony Fingerprint-Based Discrimination of Staphylococcus species with Machine Learning Approaches

• Published in: Sensors (Basel, Switzerland) Vol. 18; no. 9; p. 2789
• Main Authors: Maeda, Yoshiaki, Sugiyama, Yui, Kogiso, Atsushi, Lim, Tae-Kyu, Harada, Manabu, Yoshino, Tomoko, Matsunaga, Tadashi, Tanaka, Tsuyoshi
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 24.08.2018; MDPI
• Subjects: Artificial intelligence; Bacteria; Colonies; colony fingerprinting; Discrimination; Fingerprinting; Fingerprints; Lasers; Learning theory; lens-less imaging; machine learning; Methods; Milk; Morphology; Multivariate analysis; Phenols; Pseudomonas aeruginosa; Ribosomal DNA; Semiconductors; Staphylococcus species; Support vector machines; Japan; colony fingerprinting; Staphylococcus species; lens-less imaging; machine learning
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s18092789

Detection and discrimination of bacteria are crucial in a wide range of industries, including clinical testing, and food and beverage production. species cause various diseases, and are frequently detected in clinical specimens and food products. In particular, is well known to be the most pathogenic species. Conventional phenotypic and genotypic methods for discrimination of spp. are time-consuming and labor-intensive. To address this issue, in the present study, we applied a novel discrimination methodology called colony fingerprinting. Colony fingerprinting discriminates bacterial species based on the multivariate analysis of the images of microcolonies (referred to as colony fingerprints) with a size of up to 250 μm in diameter. The colony fingerprints were obtained via a lens-less imaging system. Profiling of the colony fingerprints of five spp. ( , , , , and ) revealed that the central regions of the colony fingerprints showed species-specific patterns. We developed 14 discriminative parameters, some of which highlight the features of the central regions, and analyzed them by several machine learning approaches. As a result, artificial neural network (ANN), support vector machine (SVM), and random forest (RF) showed high performance for discrimination of theses bacteria. Bacterial discrimination by colony fingerprinting can be performed within 11 h, on average, and therefore can cut discrimination time in half compared to conventional methods. Moreover, we also successfully demonstrated discrimination of in a mixed culture with . These results suggest that colony fingerprinting is useful for discrimination of spp.