4.3.2
IMAGE ANALYSIS AS AN AID TO DISCRIMINATION BETWEEN COLLETOTRICHUM ACUTATUM AND C. GLOEOSPORIOIDES

CR LANE1, H JEULAND2, AG HALL3 and ED CHESMORE3

1Central Science Laboratory, MAFF, Sand Hutton, York YO4 1LZ, UK; 2Ecole Nationale Superieure de Physique de Strasbourg, Strasbourg, France; 3Department of Electronic Engineering, University of Hull, Hull HU6 7RX, UK

Background and objectives
Colletotrichum acutatum, the causal agent of strawberry black spot, is an EC-listed quarantine fungus. It is distinguished from Colletotrichum gloeosporioides by the shape of the spore apex, described as acute for C. acutatum and obtuse for C. gloeosporioides [1]. The ability of image analysis to quantify this difference and hence reliably separate these morphologically similar species was studied.

Materials and methods
Images of conidia were scanned from line drawings of representative species [1] and acquired from slides prepared from cultures. Slides were viewed at 1000x magnification. Images were obtained with a three-chip colour camera (JVC-TK 1270E) and captured with a Snapper 24 frame grabber using computer software from Optimas (Seattle, USA). Discrimination of the shapes from the background (thresholding) was improved by staining with cotton blue [2]. Length (principal axis), breadth (minor axis) and area of conidia were measured using Optimas (version 5.2) and the data exported to Microsoft Excel where the area bounding the spore (length x breadth) and the area/bounding area were calculated. Images were then analysed with Matlab (version 4). A theoretical ellipse was generated from the principal and minor axes and then compared with the actual spore contour (outline). The principal and minor axes were also used to separate each spore into four segments, and individual areas calculated. These were placed in a vector and multiplied by the Hadamard matrix to calculate the Hadamard-Walsh spectrum of the spore shape (a measure of symmetry properties) [3].

Results and conclusions
As might have been expected for such small, similar and naturally variable objects, separation was difficult and required complex analysis with signal processing software (Matlab) and not the simple parameters supplied with Optimas software. However, there was a statistically significant difference between C. acutatum and C. gloeosporioides for the area/bounding area for conidial line drawings. Although a similar trend was obtained for conidia from cultures, it was not statistically significant. Comparison of spores with a theoretical ellipse failed to separate species reliably. The Hadamard function, however, showed greater promise. The vertical component for C. acutatum ranged from 1 to 2, whilst for C. gloeosporioides it was between 0 and 1, showing reasonable discrimination. This study, whilst still in its infancy, demonstrated that image analysis can quantify morphologically similar fungal structures and compare features from line drawings with material from cultures.

The first author wishes to acknowledge the financial support of the Plant Health Division of the Ministry of Agriculture, Fisheries and Food, the kind assistance of colleagues at CSL and excellent collaboration with the University of Hull.

References
1. Sutton BC, 1980. The Coelomycetes. CMI, Kew.
2. Lane CR, 1997. Proceedings of the 4th International Symposium of the EFPP, Bonn, 1996.
3. Jeuland H, 1997. Project Report, Department of Electronic Engineering, University of Hull.