Repeatability and comparison of image analysis and visual assessment for disease assessment of citrus canker

• Published in: Phytopathology Vol. 96; no. 6; pp. S13 - S14
• Main Authors: Bock, CH, Parker, P E, Cook, A Z, Gottwald, T R
• Format: Journal Article
• Language: English
• Published: 01.06.2006
• Subjects: Citrus; Xanthomonas axonopodis
• ISSN: 0031-949X

Disease assessment is required for monitoring disease either for crop management or research purposes. It must be accurate, precise and repeatable to obtain quality data. Visual assessment of the incidence and severity of citrus canker (caused by Xanthomonas axonopodis pv. citri) was compared to image analysis (IA). Digital images of 214 citrus leaves with a range of incidence and severity of citrus canker were assessed by three plant pathologists (VAs) and by IA on two occasions. Numbers of lesions/leaf (NLL) were counted, and total % area infected (TA) and % area affected by necrosis (NA) estimated. Regression analysis showed a fair relationship between IA and VAs counts of lesions in the 1st (R super(2) = 0.71-0.89) and 2nd assessment times (R super(2) = 0.78-0.90), but there was less agreement on TA and NA (R super(2) = 0.44-0.83). NLL counted by IA was 19.1 and 19.2 on the 1st and 2nd assessments, respectively, while VA counts ranged from 16.8 to 24.7. The TA measured by IA was 9.4 and 9.1%, respectively, while VA estimates ranged from 7.7 to 14.8%. For all variables IA was more repeatable (R super(2) = 0.93-0.99) compared to VAs (R super(2) = 0.42-0.93), suggesting greater consistency with the IA method. Error in IA may originate from limitations when confronted with heterogeneous symptoms or coalesced lesions, while assessors can lose concentration counting and have trouble estimating relative areas.