Identifying Fungal Leaf Spots on Oilseed Rape: When Reference Isolates, Standardised Images, Multi‐Expert Image Annotation and Molecular Diagnostics Contribute Jointly
- 29 June 2025
- journal article
- Published by Wiley in Plant Pathology
- Vol. 74 (7) , 2096-2107
Abstract
From sowing until harvest, oilseed rape is infected by several fungi, thus leaf spot symptoms coexist on plants. Training experts in identification is quick for typical leaf spots. However, for atypical spots or when colour and morphology are similar between species, there is a risk of confusion, which may preclude the identification of the causal agent. We confirmed the taxonomic assignation of a collection of isolates belonging to the main fungal oilseed rape pathogens with four barcode genes sequenced, and an in planta quantitative PCR (qPCR) diagnostic tool was developed. We produced an image database of 937 leaf spots across varieties and sampling times. A web interface developed for the independent visual diagnosis by six experts detected contrasted situations of expertise ranging from consensual identification to disagreement or lack of identification. Consensual assignation to fungal species of 82 leaf fragments was confirmed by qPCR. We then used qPCR for 101 leaf spots with a non‐consensual diagnosis. Statistical analysis of the confusion matrix confronting experts' marks to the qPCR results indicated a very good performance on typical leaf spots (accuracy = 0.88) and a fair performance on atypical leaf spots (accuracy = 0.43). Accuracy depended on the disease, highlighting those difficult to recognise like Plenodomus biglobosus , Neopseudocercosporella capsellae and Pyrenopeziza brassicae , indicating the need to adjust training. We selected images illustrating the diversity of typical and atypical leaf spots to produce diagnostic sheets. This database and the diagnostic sheets will enable methodological developments in computer vision, diagnostic assistance and expert training.Keywords
Funding Information
- Agence Nationale de la Recherche (ANR‐20‐PCPC‐0004, ANR‐17‐EUR‐0007)
- Institut National de Recherche pour l'Agriculture, l'Alimentation et l'Environnement
This publication has 27 references indexed in Scilit:
- Plant disease recognition in a low data scenario using few-shot learningComputers and Electronics in Agriculture, 2024
- Image‐based detection and classification of plant diseases using deep learning: State‐of‐the‐art reviewUrban Agriculture & Regional Food Systems, 2024
- Crop diagnostic system: A robust disease detection and management system for leafy green crops grown in an aquaponics facilityArtificial Intelligence in Agriculture, 2023
- A special issue on phytopathometry — visual assessment, remote sensing, and artificial intelligence in the twenty-first centuryTropical Plant Pathology, 2022
- Comprehensive account of present techniques for in-field plant disease diagnosisArchives of Microbiology, 2021
- Plant disease severity estimated visually: a century of research, best practices, and opportunities for improving methods and practices to maximize accuracyTropical Plant Pathology, 2021
- Virulence/avirulence patterns among Leptosphaeria maculans isolates determines expression of resistance, senescence and yellowing in cotyledons of Brassica napusEuropean Journal of Plant Pathology, 2020
- New perspectives on plant disease characterization based on deep learningComputers and Electronics in Agriculture, 2020
- Coming up short: Identifying substrate and geographic biases in fungal sequence databasesFungal Ecology, 2018
- A review on the main challenges in automatic plant disease identification based on visible range imagesBiosystems Engineering, 2016