Application of Different Classification Methods to Determine the Developmental Stage of Fish Erythrocytes of the Common Nase (Chondrostoma nasus) and Vimba Bream (Vimba vimba)

• Published in: Pattern recognition and image analysis Vol. 30; no. 1; pp. 43 - 51
• Main Authors: Spasić, Sladjana, Subotić, Srdjan, Višnjić-Jeftić, Željka, Lenhardt, Mirjana
• Format: Journal Article
• Language: English
• Published: Moscow Pleiades Publishing 2020; Springer Nature B.V
• Subjects: Bream; Cluster analysis; Computer Science; Erythrocytes; Image Processing and Computer Vision; Multilayer perceptrons; Neural networks; Nuclei (cytology); Pattern Recognition; Regression analysis; Regression models; Special Issue; Two Step Cluster analysis; K-Means Cluster Analysis; shape factor of erythrocytes; Ordinal Logistic Regression analysis; Multilayer Perceptron; developmental stage of erythrocytes
• ISSN: 1054-6618; 1555-6212
• DOI: 10.1134/S1054661820010150

Fish erythrocytes released from erythropoietic sites are immature. During circulation, they increase their own surface, while the cell body and nuclei begin to become more elliptical and less spherical during maturation. The relative abundance of different developmental stages represents an erythron profile which could be a more sensitive indicator of contamination than classic hematological indices. In this study, we analyzed blood smears of the common nase ( Chondrostoma nasus ) and vimba bream ( Vimba vimba ) with the main goal to identify the determinants of the developmental stage of erythrocytes. Based on parameters developmental stages, erythrocytes are categorized into immature, intermediate or mature. In this investigation we have used four classification methods: the Two Step Cluster analysis, the K-Means Cluster analysis, and Neural Networks–Multilayer Perceptron and an Ordinal Regression model. Our findings clearly justify that Multilayers Perceptron and OLR models are appropriate to classify the developmental stage of fish erythrocytes in both species.