www.astesj.com 347 Zebrafish Larvae Classification based on Decision Tree Model: A Comparative Analysis Bayan AlSaaidah *,1 , Waleed Al-Nuaimy 2 , Mohammed Rasoul Al-Hadidi 3 , Iain Young 4 1 Department of Electrical Engineering and Electronics, University of Liverpool, Liverpool, L69 3GJ, UK 2 Department of Electrical Engineering and Electronics, University of Liverpool, Liverpool, UK 3 Department of Computer Engineering, Al-Balqa Applied University, Al-Salt, Jordan 4 Institute of Integrative Biology, University of Liverpool, Liverpool, UK A R T I C L E I N F O A B S T R A C T Article history: Received: 18 July, 2018 Revised: 09 August, 2018 Accepted: 21 August, 2018 Screening the abnormal development of the zebrafish embryos before and after being hatched for a large number of samples is always carried out manually. The manual process is presented as a tedious work and low-throughput. The single female fish produce hundreds of eggs in every single mating process, the samples of the zebrafish embryos should be studied and analyzed within a short time according to the fast response of their bodies and the ethical regulations. The limited number of the automatic screening systems for aquaculture experiments encourage researchers to find out a high-throughput screening systems with a fast prediction results according to the large number of experimental samples. This work aims to design an automatic segmentation, classification system for zebrafish eggs using two ways for feature extraction and also a classifier. Using the whole image generally with several feature vectors useful for detection process, this way does not depend on the type of the image. The second way focus on specific characteristics of the image which are the colour and the texture features relating to the system purposes. Two different ways for feature extraction integrated by the Classification And Regression Tree (CART) classifier are proposed, analysed, and qualified by comparing the two methods performance and accuracies. The experimental results for zebrafish eggs classification into three distinct classes: live egg, live embryo, dead egg show higher accuracy using the texture and colour feature extraction with an accuracy 97% without any manual intervention. The proposed system results very promising for another type of classification such as the zebrafish larva deformations. Keywords: Classification CART Model Feature Extraction Zebrafish 1. Introduction This paper is an extended work for the published paper in International Conference on Information and Communication Systems (ICICS) [1]. The proposed procedure is a part of an integrated detection, classification, counting system for zebrafish embryo malformations. After adding different chemical substances with different concentration, several deformation types appear on the larva body whether before or after being hatched. The malformations classified depending on the affected part such as the tail curvature, necrosed yolk, and the dead larva. Over the recent years the zebrafish has become one of the most common animal models. This is due to many factors including a high degree of genetic similarity with humans, short generation times, transparent larval stages, extensively annotated genome and simple husbandry [2][3]. Zebrafish are now widely used in drug development, to measure the impact of environmental changes, of toxins and pollutants and many other applications. However, the use mammals in the biological experiment is expensive and laborious, it also led to an increasing number of ASTESJ ISSN: 2415-6698 * Bayan AlSaaidah, University of Liverpool, Email: bayan@liv.ac.uk Advances in Science, Technology and Engineering Systems Journal Vol. 3, No. 4, 347-353 (2018) www.astesj.com Special Issue on Recent Advances in Engineering Systems https://dx.doi.org/10.25046/aj030435