Near infrared hyperspectral imaging : a rapid method for the differentiation of maize ear rot pathogens on growth media

Bezuidenhout, Cenette (2018-03)

Thesis (MSc Food Sc)--Stellenbosch University, 2018.

Thesis

ENGLISH ABSTRACT: Maize grain is highly susceptible to the toxin-producing fungal pathogens Fusarium spp. and Stenocarpella spp. Infection of grain with these species leads to maize ear rots but of greater concern is their ability to produce mycotoxins, which can promote cancer, among other diseases, in humans and animals. By combining microbiology, plant pathology and chemistry disciplines to create rapid screening methods that can accurately distinguish fungal pathogens, an initial step in ensuring food safety can be achieved, which can potentially be applied to maize grain in future. This thesis aimed to distinguish between the most important maize ear rot pathogens namely Fusarium verticillioides, F. graminearum species complex (FGSC) and Stenocarpella maydis that cause Fusarium-, Gibberella- and Diplodia ear rot, respectively. Furthermore, pathogen isolates from the same species were also distinguished. This was done with near infrared (NIR) hyperspectral imaging, a technology that offers the ability for rapid sample measurement that provides data containing both spatial and spectral information. Through multivariate analysis such as principal component analysis, primarily used for data exploration, and partial least square discriminant analysis, hyperspectral images was used to build classification models that accurately distinguish between the maize ear rot pathogens. All the major ear rot pathogens were distinguished with reasonable accuracy on day 3 of growth, with the isolates from the same species showing higher accuracy on day 5 of growth.

AFRIKAANSE OPSOMMING: Mieliegraan is hoogs vatbaar vir die toksien-produserende swam-patogene naamlik Fusarium spp. en Stenocarpella spp. Infeksie van graan met hierdie spesies lei tot mielie-kopvrot, en wat groter kommer wek, is dat hulle oor die vermoë beskik om mikotoksiene te produseer, wat siektes soos kanker by mense en diere bevorder. Deur die kombinasie van mikrobiologie, plantpatologie en chemie dissiplines om vinnige siftingsmetodes te skep wat die swam-patogene akkuraat kan onderskei, kan 'n aanvanklike stap in die versekering van voedselveiligheid bereik word, wat moontlik in die toekoms op mieliegraan toegepas kan word. Hierdie proefskrif het ten doel om te onderskei tussen die belangrikste mielie-kopvrot-patogene, naamlik Fusarium verticillioides, Fusarium graminearum spesie kompleks (FGSK) en Stenocarpella maydis wat onderskeidelik Fusarium-, Gibberella- en Diplodia- kopvrot veroorsaak. Verder is patogeen-isolate van dieselfde spesie ook onderskei. Dit is gedoen met naby infrarooi (NIR) hiperspektrale beelding, 'n tegnologie wat die vermoë bied vir vinnige monstermeting wat data bevat wat beide ruimtelike en spektrale inligting bevat. Deur middel van meerveranderlike-beeldontleding soos hoofkomponent-analise, hoofsaaklik vir dataverkenning, en pasiële kleinste waarde diskriminant-analise, is hiperspektrale beelde gebruik om klassifikasie modelle te bou wat akkuraat tussen die mielie-kopvrot-patogene onderskei. Al die hoof kopvrot-patogene is met redelike akkuraatheid op dag 3 van groei onderskei, met die isolate van dieselfde spesie wat hoër akkuraatheid op dag 5 van groei toon.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/103397
This item appears in the following collections: