You are on page 1of 4

A PRELIMINARY STUDY FOR CONSTRUCTING A COMPUTATIONAL PROCEDURE FOR NEMATODES IDENTIFICATION BASED ON MORPHOLOGICAL ASPECTS

Luis A.M. Palhares de Melo, 2Rogerio C. Bastos

1CENARGEN - EMBRAPA - Centro Nacional de Recursos Genticos e Biotecnologia, SAIN Parque Rural Final W5 Norte, 70.770-910, BrasliaDF, BRASIL, e-mail: palhares@cenargen.embrapa.br, 2UFSC Universidade Federal de Santa Catarina, Ps-Graduao em Cincia da Computao, Campus Trindade, 88.040-900, Florianpolis-SC, BRASIL e-mail: rogerio@eps.ufsc.br

Abstract: Nematodes are vermiform animals of long and thin body and slender at the two extremities. They have great economic importance because many are plant parasitic and very difficult to erradicate once established in the soil. A prerequisite to combating their agricultural impact is correct identification of species. We present a preliminary study for constructing a computational procedure for nematode identification (in terms of taxonomic classification) based on morphological aspects of various organs (stylet, ovary, tail, esophagus, etc). The procedure is an artificial intelligent application involving concepts of computer vision and pattern recognition (an automated system that "visually recognizes nematode's structures). Keywords: Computer vision; Pattern recognition; Nematode

1 Characterizing nematodes Nematodes are vermiform animals of long and thin body and slender on the two extremities, that are found everywhere in the planet where life can exist. Beside their typically cylindrical format, nematodes present a complete digestive duct with mouth and anus, and bilateral symmetry. Various organs compose nematode's body including stylet, metacorpus, isthmus, ovary, esophagus and annulis.

Figure 1. Drawings of males and second-stage juveniles of Meloidogyne paranaensis n. sp. A) Anterior portion of male. B) Male tail, lateral view. (Carneiro (1996)) Many kinds of nematodes attack most plants of economic importance. Different species cause different types of damage, for instance, some cause seed destruction, others a drastic reduction of the

First European Conference for Information Technology in Agriculture, Copenhagen, 1518 June, 1997

productivity and quality of planted land. The first step in reducing nematode damage is the correct diagnosis of the problem and of the nematode causing it.

2 Development of computer identification programs Boag et al. (1988) comment that most nematode species are identified using a number of morphometric and morphological characters and the morphometric characters are particularly suitable for measurement and analysis by a computer. Also, they point out some computer identification programs based on the nematode's morphometric characters, have already been written for identifying genus Helycotilenchus and Longidorus. Peet et al. (1990) developed a digital imaging system for measuring various physical characteristics of individual nematodes and for comparing groups of nematodes. La Blanca et al. (1992) and Valdivia et al. (1992) developed computational procedures for estimating and calibrating nematode features from digital images. As they point out, feature identification from digitized microscopic images has been used for more than 10 years but is relatively new in nematology. Bravo & Roca (1994) wrote a program for identifying genus Xiphinema using dBASE IV software and Robbins & Brown (1994) did the same to genus Longidorus using LOTUS.

3 A computational procedure for identifying nematodes 3.1 The main steps to proceed identification To process morphological (and morphometrical) characters automatically we have to take into account aspects of computer vision and pattern recognition where the source data to be initially processed are pictures contained in digital images. We started a preliminary study with this idea in mind and developed a software prototype (using C++ language) that tries to recognize nematodes (in terms of taxonomy) based only on stylets morphology aspect. We assumed that only stylets shape is sufficient to correctly identify the nematode although we know that this is not the case. The basic idea is recognition based on template matching technique. A database was built using contour sketches of stylets of nematodes of known taxonomy. Having an unknown nematode digitized image, the computational procedure must localize the stylet, and constructs a sketch (the morphological aspect considered). Then by template matching, based on fuzzy sets (Zadeh (1965), Shaout & Suk (1992)) the procedure compares these sketch against all stylets sketches stored previously on the database. A distance (similarity measure) ranging from zero to one is generated for every comparison. A distance zero indicates that sketchs are strictly identical, and a distance one that they are completely different from each other. In figure 2 we see the results produced by the program after storing a new stylets shape sketch in the database. Initially an unknown digitized nematode image focusing stylet is presented to the program (window NEMAT07.BMP). The program then performs image segmentation generating a window (SEGMENT.BMP). We have used a generic segmentation algorithm proposed by Horowitz & Pavlidis (1976) but other segmentation techiniques could and must be implemented to improve segmentation performance. The automated procedure that locates the stylets position was not considered at this stage. Therefore, to locate the stylet in digitized images, the programs operator points out the region corresponding to it with the help of mouse in window SEGMENT.BMP. So the program isolates stylet and shows it in window ESTILETE.BMP. Once the stylet is isolated, the program produces its sketch based on an very simple algorithm of polygonal approximation proposed by Pavilids & Horowitz (1974) that must be further complemeted by others that improve the final layout sketch. Finally, the window Identificar Estilete registers generic information (including taxonomic propreties) about the neamtode. These data and the stylets sketch are stored in the database.

Figure 2. Results of entire process of registering a new nematode in the database. 3.2 What must the computational procedure anserws ? We have implemented a single test to evaluate the feasibility of the process. In our test we stored in the database three stylets sketches of three distinct nematodes. They were stored in files TEMPL01, TEMPL02 and TEMPL03. In file PADREC was stored a stylet's sketch of an unknown nematode. The test consists of calculating the distance measure (based on fuzzy sets) between PADREC and TEMPL01, TEMPL02 and TEMPL03 respectively (shown in figure 3).

Figure 3. Stylets sketches used in the test of the program.

Figure 4. Template matching between PADREC and TEMPL01, TEMPL02 and TEMPL03 The question the program answers is: Which of the three pictures (TEMPL01, TEMPL02, TEMPL03) has more resemblance with PADREC ?. The distance between PADREC and TEMPL01, TEMPL02 and TEMPL03 calculated by the program is respectively 0.261199, 0.351007 and

0.364595 (see figure 4). PADREC will be assumed as having the features of nematode with stylet's sketch TEMPL01 (0.261199, the least distance). Yet in figure 4 we can see that numeric distance between PADREC and TEMPL02 and between PADREC and TEMPL03 are too close (0.351007 and 0.364595) and it does not reflect semantically the situation. That is, if distance between PADREC and TEMPL02 is quite the same as distance between PADREC and TEMPL03 it would be expected that TEMPL02 and TEMPL03 are closely identical and that's not the case. To manage this problem, we must forward in our study, redefine the fuzzy sets and general conditions that support our template matching scheme.

4 Conclusions With this preliminary study we have shown that a computational procedure for identifying nematodes based on aspects of computer vision and pattern recognition is feasible. Only morphological aspects were taken into account in this work. We later intend to improve the power of the computational procedure, considering morphometric aspects, as well as the management of spatial relations between organs, which compound some nematode identification key, like "esophagus overlaps intestine ventrally", "procorpus and metacorpus not swollen and combined into a large valvular bulb", "lateral field with 4 incisures", among others.

5 References Boag, B., P.B.Tophan, D.J.F. Brown & P. Smith (1988). The use of micro computers for the identification of plant-parasitic nematodes. In: Nematode Identification and Expert System Technology. 9-15. Plenum Press, New York, London. Bravo, M.A. & F. Roca (1994). Using dBASE for Identification of Species of the Genus Xiphinema. In: 22nd International Symposium Proceedings of the European Society of Nematologists, Ghent, Belgium, Aug/1994. Carneiro, R.M.D.G., R.G. Carneiro, I.M.O. Abrantes, M.S.N.A. Santos & M.R.A. Almeida (1996). Meloidogyne paranaensis n. sp. (Nemata: Meloidogynidae), A Root-Knot Nematode Parasitizing Coffee in Brazil. In: Journal of Nematology 28 (2): 177-189 Horowitz, S.L. & T. Pavlidis (1976). Picture Segmentation by a Tree Traversal Algorithm. In: Journal of the Association for Computing Machinery 23 (2): 368-388 La Blanca, N.P., J.F. Valdivia, P. Castillo & A.G. Barcina (1992). Detecting Nematode Features from Digital Images. In: Journal of Nematology 24 (2) : 289-297. Melo, L.A.M.P. (1996). A Model for Nematode Identification based on Stylet Structure. Msc Thesis. Federal University of Santa Catarina- Brazil. (In portuguese) Pavlidis, T. & S.L. Horowitz (1974). Segmentation of Plane Curves. In: IEEE Transactions on Electronic Computers 23 (8) : 860-870 Peet, F.G., T.S. Panesar, T.S. Sahota & J.R. Sutherland (1990). A Digital Image Analysis System for Comparing Groups of Small Nematodes. In: Journal of Nematology 22(3) : 407-413. Robbins, R.T. & D.J.F. Brown (1994). A Computer Program "LONG-SORT" to Assist with the Diagnosis and Identification of longidorus Species. In: 22nd International Symposium Proceedings of the European Society of Nematologists, Ghent, Belgium, Aug/1994. Shaout, A. & M. Suk (1992). Distance Measure for Attributed Fuzzy Tournaments. In: IEE Proceedings-E 139 (5) : 373-378. Valdivia, J.F., N.P. La Blanca, P. Castillo & A.G. Barcina (1992). Line Detection and Texture Analysis for Automatic Nematode Identification. In: Journal of Nematology 24 (4) : 571-577. Zadeh, L.A. (1965). Fuzzy Sets. In : Information and Control 8 : 338-353

You might also like