Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging
Thesis (MScFoodSc)--Stellenbosch University, 2023.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613878013132800 |
|---|---|
| access_status_str | Open Access |
| author | Mannix, Kim Teresa |
| author2 | Williams, Paul James |
| author_browse | Mannix, Kim Teresa Williams, Paul James |
| author_facet | Williams, Paul James Mannix, Kim Teresa |
| author_sort | Mannix, Kim Teresa |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MScFoodSc)--Stellenbosch University, 2023. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/128448 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:43:08.148Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/128448 Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging Mannix, Kim Teresa Williams, Paul James Rose, Lindy J. Stellenbosch University. Faculty of AgriSciences. Dept. of Food Science. Corn -- Seeds Mycotoxins Toxigenic fungi Fusarium diseases of plants -- Detection Corn -- Diseases and pests -- Biological control -- Economic aspects Plants -- Effect of mycotoxins on Spectral imaging -- Data processing UCTD Thesis (MScFoodSc)--Stellenbosch University, 2023. ENGLISH ABSTRACT: Maize is of great importance for human and animal consumption on both a global and local scale. Plant diseases caused by mycotoxigenic fungi result in economic and yield losses. More significant, however, is the production of mycotoxins by these micro-organisms that poses a prominent food safety risk. Spectral imaging has been explored for the early detection of mycotoxigenic fungi in maize whereby several considerations have been previously investigated. Early detection studies typically use spore inoculum as the infection source. However, several pathways of infection exist which may originate from either spores or mycelia. Therefore, the study objectives were to determine whether there was a spectral difference between the type of inoculum (spores vs mycelia); and to investigate the effect of inoculum concentration on infection source discrimination. Maize kernels were artificially inoculated with Fusarium verticillioides inoculum: either mycelial suspension (mycelia) or spore suspension (spores). Kernels inoculated with distilled water served as the control. These kernels were imaged daily for four days post-inoculation. Multiclass and binary classification models were developed using five different classification algorithms. Multiclass classification evaluated three classes namely: control, mycelia, and spores. On the other hand, binary classifications evaluated two classes namely control and infected (this class merged the mycelia class with the spore class to form one infected class). PCA-LDA, PLS-DA, and SVM models generally performed the best. Binary classification performed superiorly to multiclass classifications. Classification accuracy ranged from 73.3% to 96.7% from day 1 to day 4 for binary classification. A confirmation study investigating day 1 post-inoculation showed that results of multiclass (74.7%) and binary (86.7%) classification were improved upon the addition of more samples, although a high degree of variation between cross-validation folds was noted. Varying concentrations (low, mid, and high) of mycelial inoculum were individually compared to spore and control classes to assess how inoculum concentration influences classification ability. High (74.7% for multiclass and 86.7% for binary) and mid (77.3% for multiclass and 84.7% for binary) concentrations produced comparable results. The low concentration showed a decrease in classification accuracy (70.0% for multiclass and binary) compared to previous groups. Large variation was noted in the results of the confirmation study in the first objective and second objective. To conclude, analysis over time suggests no difference between inoculum sources after day 3 post-inoculation. The initial period of infection (day 1) shows an improved ability to distinguish between inoculum sources with the addition of more samples, although variation also increases. During the initial period of infection (day 1), mycelial inoculum may affect the classification ability of the model. This study has thus expanded on some considerations for the early detection on maize using spectral imaging. AFRIKAANSE OPSOMMING: Mielies is van groot belang vir beide menslike en dierlike verbruik op globale en plaaslike skaal. Plantsiektes wat deur mikotoksiese swamme veroorsaak word lei tot ekonomiese- en opbrengsverliese. Meer noemenswaardig is egter die produksie van mikotosiene deur mikroörganismes wat ‘n prominente voedselveiligheidsrisiko inhou. Spektrale beelding is ondersoek vir die vroeë opsporing van mikotoksiese swamme in mielies, waardeur verskeie oorwegings voorheen ondersoek is. Vroeë opspooringstudies gebruik tipies spoor-inokulums as infeksiebron. Infeksie kan egter ontstaan vanaf spore of miselium. Daarom was die doelwitte van die studie om te bepaal of daar ‘n spektrale verskil tussen die tipe inokulum (spore vs miselium) was; en om die effek van inokulumkonsentrasie op infeksiebron-diskriminasie te ondersoek. Mieliepitte is kunsmatig met Fusarium verticilliodes inokulum geënt: óf miselium-suspensie (miselium) óf spoor-suspensie (spore). Mieliepitte wat met gedistilleerde wat geënt is, het as kontrole gedien. Hierdie mieliepitte is daagliks vir vier dae na inokulasie afgebeeld. Multiklas- en binêre klassifikasiemodelle is ontwikkel met behulp van vyf verskillende klassifikasie-algoritmes. Multiklasklassifikasie het drie klasse geëvalueer, naamlik: kontrole, miselium en spore. Aan die ander kant het binêre klassifikasies twee klasse geëvalueer, naamlik kontrole en geïnfekteerde (hierdie klas het die miseliumklas met die spoorklas saamgesmelt om een geïnfekteerde klas te vorm). PCA-LDA-, PLS-DA- en SVM-modelle het oor die algemeen die beste gevaar. Binêre klassifikasie het beter presteer as multiklas klassifikasies. Klassifikasie akkuraatheid het gewissel van 73.3% tot 96.7% vanaf dag 1 tot dag 4 vir binêre klassifikasie. 'n Validasiestudie wat dag 1 na-inokulasie ondersoek het, het getoon dat resultate van multiklas (74.7%) en binêre (86.7%) klassifikasie verbeter is met die byvoeging van meer monsters, alhoewel 'n hoë mate van variasie tussen kruisvalideringsvoue opgemerk is. Verskillende konsentrasies (laag, middel en hoog) van miselium inokulum is individueel vergelyk met spoor- en kontroleklasse om te bepaal hoe inokulumkonsentrasie klassifikasievermoë beïnvloed. Hoë (74.7% vir multiklas en 86.7% vir binêre) en middel (77.3% vir multiklas en 84.7% vir binêre) konsentrasies het vergelykbare resultate gelewer. Die lae konsentrasie het 'n afname in klassifikasie akkuraatheid (70.0% vir multiklas en binêre) getoon in vergelyking met vorige groepe. Groot variasie is opgemerk in die resultate van die validasiestudie in die eerste en tweede doelwitte. Ter afsluiting: analise oor tyd dui op geen verskil tussen inokulumbronne na dag 3 na-inokulasie nie. Die aanvanklike tydperk van infeksie (dag 1) toon 'n verbeterde vermoë om tussen inokulumbronne te onderskei met die byvoeging van meer monsters, alhoewel variasie ook toeneem. Gedurende die aanvanklike tydperk van infeksie (dag 1), kan miselium inokulum die klassifikasievermoë van die model beïnvloed. Hierdie studie het uitgebrei oor sommige oorwegings vir die vroeë opsporing op mielies met behulp van spektrale beelding. Masters 2023-03-04T09:02:31Z 2023-08-30T13:07:49Z 2023-03-04 2023-03-04T09:02:31Z 2023-08-31T09:18:45Z 2023-03-04T09:02:31Z 2023-08-31T09:18:45Z 2023-03 Thesis https://scholar.sun.ac.za/handle/10019.1/128448 en Stellenbosch University application/pdf xv, 109 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Corn -- Seeds Mycotoxins Toxigenic fungi Fusarium diseases of plants -- Detection Corn -- Diseases and pests -- Biological control -- Economic aspects Plants -- Effect of mycotoxins on Spectral imaging -- Data processing UCTD Mannix, Kim Teresa Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| title | Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| title_full | Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| title_fullStr | Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| title_full_unstemmed | Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| title_short | Considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| title_sort | considerations for the early detection of mycotoxigenic fungi on maize kernels using spectral imaging |
| topic | Corn -- Seeds Mycotoxins Toxigenic fungi Fusarium diseases of plants -- Detection Corn -- Diseases and pests -- Biological control -- Economic aspects Plants -- Effect of mycotoxins on Spectral imaging -- Data processing UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/128448 |
| work_keys_str_mv | AT mannixkimteresa considerationsfortheearlydetectionofmycotoxigenicfungionmaizekernelsusingspectralimaging |