Identification of Selected Monogeneans Using Image Processing, Artificial Neural Network and K-Nearest Neighbor

Identification of selected monogeneans using image processing, artificial neural network and K-nearest neighbor Item Type article Authors Yousef Kalafi, E.; Wooi Boon, T.; Town, C.; Kaur Dhillon, S. Download date 02/10/2021 11:21:57 Link to Item http://hdl.handle.net/1834/40338 Iranian Journal of Fisheries Sciences 17(4) 805-820 2018 DOI:10.22092/ijfs.2018.117017 Identification of selected monogeneans using image processing, artificial neural network and K-nearest neighbor ٭Yousef Kalafi E.¹; Wooi Boon T.¹; Town C.²; Kaur Dhillon S.¹ Received: June 2017 Accepted: August 2017 Abstract Over the last two decades, improvements in developing computational tools have made significant contributions to the classification of images of biological specimens to their corresponding species. These days, identification of biological species is much easier for taxonomists and even non-taxonomists due to the development of automated computer techniques and systems. In this study, we developed a fully automated identification model for monogenean images based on the shape characters of the haptoral organs of eight species: Sinodiplectanotrema malayanum, Diplectanum jaculator, Trianchoratus pahangensis, Trianchoratus lonianchoratus, Trianchoratus malayensis, Metahaliotrema ypsilocleithru, Metahaliotrema mizellei and Metahaliotrema similis. Linear Discriminant Analysis (LDA) method was used to reduce the dimension of extracted feature vectors which were then used in the classification with K-Nearest Neighbor (KNN) and Artificial Neural Network (ANN) classifiers for the identification of monogenean specimens of eight species. The need for the discovery of new characters for identification of species has been Downloaded from jifro.ir at 12:07 +0330 on Saturday October 6th 2018 acknowledged for log by systematic parasitology. Using the overall form of anchors and bars for extraction of features led to acceptable results in automated classification of monogeneans. To date, this is the first fully automated identification model for monogeneans with an accuracy of 86.25% using KNN and 93.1% using ANN. Keywords: Monogenean, Morphology, Fish parasite, Artificial neural networks, K- nearest neighbor 1-Institute of Biological Sciences, Faculty of Science, University of Malaya, Kuala Lumpur, Malaysia 2-Computer Laboratory, University of Cambridge, Cambridge CB3 0FD, UK *Corresponding author's Email: [email protected] 806 Yousef Kalafi et al., Identification of selected monogeneans using image… Introduction morphology and ecology and with Monogeneans are platyhelminthes respect to the variation of structural which are characterized by having a designs in the attachment organs proper body and haptor with sizes (Kearn, 1994), which are usually used ranging from 0.5mm to 1-2cm in length for species identification. The haptoral live on lower aquatic invertebrates or attachment organs of monogeneans are the gills, skin or fins of fishes as hosts. sclerotized structures of anchors, bars Their appendage attachments in their and marginal hooks. In particular, the anterior and posterior (haptoral) regions morphology of each of these organsis are used to prevent physical unique to monogenean species (Boeger dislodgement from the host (Fig. 1). and Kritsky, 1993) and is used as a diagnostic feature in their taxonomical classification (Vignon, 2011). Earlier, Active Shape Models (ASM) (Ali et al., 2012) were used to classify several Gyrodactylus species according to attachment hooks. ASM were applied to extract diagnostic information from hook images as features. Extracted features were used as input data to Linear Discriminant Analysis (LDA) , K-Nearest Neighbor (KNN), Multilayer Perceptron (MLP) and Support Vector Machine (SVM) classifiers. According to Khang et al. (2016), data from size Downloaded from jifro.ir at 12:07 +0330 on Saturday October 6th 2018 and shape of anchors were generated using geometric morphometrics. They used principal components and cluster analysis to classify 13 species of Ligophorus. Innovations in the area of computer vision have significantly contributed to the development of automated taxonomic identification systems such Figure 1: Illustration of a monogenean worm as an automated identification system consisting of three main parts: head, body which estimates densities of whiteflies, and haptor. aphids and thrips in a greenhouse (CHO Monogeneans are a diverse group, with et al., 2008), automatic image several thousand species described in recognition and diagnosis of protozoan the world (Poulin, 2002). The diversity parasites (Castañón et al., 2007), of monogeneans is not only in terms of automatic recognition of biological numbers but also in terms of their particles in microscopic images Iranian Journal of Fisheries Sciences 17(4) 2018 807 (Ranzato et al., 2007), automatic methods are performed on every single detection of malaria parasites for image (specimen) which substantially estimating parasitemia (Savkare and slows down the process of Narote, 2011), automated identification identification and classification. Hence, of copepods using digital image we propose a fully automated processing and artificial neural identification model for monogeneans networks (Leow et al., 2015), which is robust with respect to variable automated identification of fish species imaging conditions and damaged based on otolith contour using short- specimens. time Fourier transform and discriminant analysis (STFT-DA) (Salimi et al., Materials and methods 2016), and other systems (Larios et al., Recognition of monogeneans is based 2008; Vogt et al., 2009; Mansoor et al., on morphometric features of their hard 2011; Feng et al., 2016; Perre et al., parts (Lim and Gibson, 2010). For this 2016). study, images of the hard haptoral Many classification methods such as organs such as anchors and bars were Artificial Neural Network (ANN) captured using a Leica digital camera (Yang et al., 2001; Cho et al., 2008; DFC 320 attached to Leica DMRB Mansoor et al., 2011), KNN (Keller et microscope at 40× magnification. The al., 1985; Parisi-Baradad et al., 2010), resolution of the images was 1044×772 SVM (Thiel et al., 1996; Pronobis et pixels and they were saved in Tagged al., 2010), Discriminant Analysis (DA) Image File format (TIF). (Thiel et al., 1996; Salimi et al., 2016), Our database consists of 160 images Decision trees (Jalba et al., 2005), from 8 species (20 images of each Semantically-Related Visual (SRV) species): Sinodiplectanotrema Downloaded from jifro.ir at 12:07 +0330 on Saturday October 6th 2018 (Feng and Bhanu, 2013), and malayanum, Diplectanum jaculator, Convolutional Neural Networks Trianchoratus pahangensis, (Gomez et al., 2016), etc. have been Trianchoratus lonianchoratus, utilized for developing automated Trianchoratus malayensis, identification systems. Metahaliotrema ypsilocleithru, Automated classification of images Metahaliotrema mizellei and of specimens requires development of Metahaliotrema similis. Fig. 2 models and methods that are able to illustrates the flowchart for the characterize species images based on development of automated the texture or shape of objects to extract identification for monogeneans. important visual information for classification. Current approaches in monogenean identification rely heavily on manual input during image processing and feature extraction such as specifying morphological landmark features. These manual identification 808 Yousef Kalafi et al., Identification of selected monogeneans using image… features for the next process in feature extraction. Preprocessing started with converting RGB images to intensity images. Then intensity images were filtered and edges of anchors and bars were detected (Fig. 4). Figure 4: The process of detecting edges from intensity image. Figure 2: Flowchart for development of proposed model for monogenean identification. Since the segmented images contained negative and positive values, the images Preprocessing were binarized with a threshold of zero. Monogenean specimen images are very Then the borders were cleared and complex due to their messy background objects smaller than 1000 pixels were and overlapping of anchors and bars. removed (Fig. 5). Coordinates of Despite consistent efforts to acquire contour pixels for species` anchors clear images, some overlapping and were calculated. Features were clutter were unavoidable (Fig. 3). extracted either from all anchors and Downloaded from jifro.ir at 12:07 +0330 on Saturday October 6th 2018 bars as a consolidated object or from individual anchors. Figure 3: The illustration of anchors and bars of Metahaliotrema ypsilocleithrum. a) The illustration of dorsal and ventral Figure 5: The process of converting binary anchors and bars. b) The microscopic image image to segmented image. of anchors and bars and their overlapping. Feature extraction Hence, preprocessing played an Binary images were used two times for important role to omit redundant feature extraction; (i) firstly using all information and to highlight reliable Iranian Journal of Fisheries Sciences 17(4) 2018 809 anchors and bars as a consolidated of 160×7 feature vector. object and (ii) secondly by calculating coordinates of one anchor and then Classification extracting the features from that anchor. The features that were extracted and Features extracted consisted of: length selected in the previous stage were used of bounding box, width of bounding as inputs to KNN and ANN classifiers box, center of bounding box, to train the system based on a training orientation

Load more