The contamination of foods and beverages by fungi is a severe health hazard. The rapid identification of fungi species in contaminated goods is important to avoid further contamination. To this end, we developed a fungal discrimination method based on the bioimage informatics approach of colony fingerprinting. This method involves imaging and visualizing microbial colonies (referred to as colony fingerprints) using a lens-less imaging system. Subsequently, the quantitative image features were extracted as discriminative parameters and subjected to analysis using machine learning approaches. Colony fingerprinting has been previously found to be a promising approach to discriminate bacteria. In the present proof-of-concept study, we tested whether this method is also useful for fungal discrimination. As a result, 5 fungi belonging to the Aspergillus, Penicilium, Eurotium, Alternaria, and Fusarium genera were successfully discriminated based on the extracted parameters, including the number of hyphae and their branches, and their intensity distributions on the images. The discrimination of 6 closely-related Aspergillus spp. was also demonstrated using additional parameters. The cultivation time required to generate the fungal colonies with a sufficient size for colony fingerprinting was less than 48 h, shorter than those for other discrimination methods, including MALDI-TOF-MS. In addition, colony fingerprinting did not require any cumbersome pre-treatment steps prior to discrimination. Colony fingerprinting is promising for the rapid and easy discrimination of fungi for use in the ensuring the safety of food manufacturing.
Keywords: Bioimage informatics; Colony fingerprinting; Fungi; Hyphae; Lens-less imaging; Machine learning.
Copyright © 2019 Elsevier B.V. All rights reserved.