Automated image processing framework helps to analyse the density of fruiting bodies of Leptosphaeria maculans on oilseedrape stems

Understanding the transmission of plant pathogen inoculum during the periods when the host plants are not present is crucial for predicting the initiation of epidemics and optimizing mitigation strategies. However, inoculum production at the end of the cropping season, survival during the intercrop period, and the emergence or release of inoculum can be highly variable, difficult to assess, and generally inferred indirectly from symptom data. As a result, a lack of large datasets hampers the study of these epidemiological processes. Here, inoculum production was studied in Leptosphaeria maculans, the cause of phoma stem canker of oilseed rape. The fungus survives on stubble left in the field, from which ascospores are released at the beginning of the next cropping season. An image processing framework was developed to estimate the density of fruiting bodies produced on stem pieces following incubation in field conditions, and a quality assessment of the processing chain was performed. A total of 2540 standardized RGB digital images of stems were then analysed, collected from 27 oilseed rape fields in Brittany over four cropping seasons. Manual post-processing removed 16% of the pictures, e.g. when moisture-induced darkening of the oilseed rape stems caused overestimation of the area covered with fruiting bodies. The potential level of inoculum increased with increasing phoma stem canker severity at harvest, and depended on the source field and the cropping season. This work shows how image-based phenotyping generates high-throughput disease data, opening up the prospect of substantially increased precision in epidemiological studies.