The use of ASM feature extraction and machine learning for the discrimination of members of the fish ectoparasite genus gyrodactylus

• Main Authors: Ali, Rozniza, Hussain, Amir, Bron, James, Shinn, Andrew
• Other Authors: Huang, T, Zeng, Z, Li, C, Leung, CS, University of Stirling, Computing Science, Institute of Aquaculture, orcid:0000-0002-8080-082X, orcid:0000-0003-3544-0519, orcid:0000-0002-5434-2685
• Format: Book Part
• Language: English
• Published: Springer 2012
• Subjects: Attachment hooks; image processing; SEM; parasite; machine learning classifier; Atlantic salmon
• Online Access: http://hdl.handle.net/1893/16513; https://doi.org/10.1007/978-3-642-34478-7_32; http://link.springer.com/chapter/10.1007/978-3-642-34478-7_32#; http://dspace.stir.ac.uk/bitstream/1893/16513/1/Use%20of%20ASM%20Feature%20Extraction%20and%20Machine.pdf

Active Shape Models (ASM) are applied to the attachment hooks of several species of Gyrodactylus, including the notifiable pathogen G. salaris, to classify each species to their true species type. ASM is used as a feature extraction tool to select information from hook images that can be used as input data into trained classifiers. Linear (i.e. LDA and KNN) and non-linear (i.e. MLP and SVM) models are used to classify Gyrodactylus species. Species of Gyrodactylus, ectoparasitic monogenetic flukes of fish, are difficult to discriminate and identify on morphology alone and their speciation currently requires taxonomic expertise. The current exercise sets out to confidently classify species, which in this example includes a species which is notifiable pathogen of Atlantic salmon, to their true class with a high degree of accuracy. The findings from the current exercise demonstrates that data subsequently imported into a K-NN classifier, outperforms several other methods of classification (i.e. LDA, MLP and SVM) that were assessed, with an average classification accuracy of 98.75%.