Automated toxicity test model based on a bio-inspired technique and AdaBoost classifier

Abstract Measuring toxicity is one of the most important steps to develop a new drug. During the drug development, animals are widely used to investigate the toxic effects by exposing them to the toxicants. Zebrafish embryo is one of the most suitable animals for testing toxicity of compounds due to (1) the transparency of zebrafish animals and (2) the production of a large number of embryos in each mating. However, due to the high number of embryos, manual inspection is not feasible enough, slow, and inaccurate. In this paper, using machine learning and bio-inspired techniques, a fully automated method was suggested to investigate the toxicity using microscope images of treated zebrafish embryos. In this method, firstly, the Segmentation-Based Fractal Texture Analysis (SFTA) technique was employed for extracting features from embryos’ images. Then, a new version of Grey Wolf Optimization (GWO) was proposed and applied to select the most discriminative features to increase the classification performance while reducing the required computational time for the classification process. Finally, the AdaBoost classifier was used to classify an unknown image to alive or coagulant (i.e. dead embryo due to its exposure to a toxic compound). The experimental results showed that the selected features using the proposed optimization algorithm achieved the highest accuracy reached to 99.47%, the maximum average reduction rate and lowest computational time. These promising results represent a good step towards using machine learning techniques along with the new version of GWO to develop a fully automated toxicity test using zebrafish embryos images.

[1]  Aboul Ella Hassanien,et al.  Towards an Automated Zebrafish-based Toxicity Test Model Using Machine Learning , 2015 .

[2]  Markus Reischl,et al.  Robust identification of coagulated zebrafish eggs using image processing and classification techniques , 2009 .

[3]  Aboul Ella Hassanien,et al.  Classification of toxicity effects of biotransformed hepatic drugs using whale optimized support vector machines , 2017, J. Biomed. Informatics.

[4]  U. Kammann,et al.  Toxic Effects of Brominated Indoles and Phenols on Zebrafish Embryos , 2006, Archives of environmental contamination and toxicology.

[5]  Keinosuke Fukunaga,et al.  Introduction to Statistical Pattern Recognition , 1972 .

[6]  Cheng-Lung Huang,et al.  A distributed PSO-SVM hybrid system with feature selection and parameter optimization , 2008, Appl. Soft Comput..

[7]  Agma J. M. Traina,et al.  An Efficient Algorithm for Fractal Analysis of Textures , 2012, 2012 25th SIBGRAPI Conference on Graphics, Patterns and Images.

[8]  N Otsu,et al.  An automatic threshold selection method based on discriminate and least squares criteria , 1979 .

[9]  Václav Snásel,et al.  Biometric cattle identification approach based on Weber's Local Descriptor and AdaBoost classifier , 2016, Comput. Electron. Agric..

[10]  Nitesh V. Chawla,et al.  SMOTE: Synthetic Minority Over-sampling Technique , 2002, J. Artif. Intell. Res..

[11]  Andrew Lewis,et al.  Grey Wolf Optimizer , 2014, Adv. Eng. Softw..

[12]  Haibo He,et al.  Learning from Imbalanced Data , 2009, IEEE Transactions on Knowledge and Data Engineering.

[13]  Daniel Neagu,et al.  Hybrid intelligent systems for predictive toxicology - a distributed approach , 2005, 5th International Conference on Intelligent Systems Design and Applications (ISDA'05).

[14]  Aboul Ella Hassanien,et al.  A Predictive Model for Toxicity Effects Assessment of Biotransformed Hepatic Drugs Using Iterative Sampling Method , 2016, Scientific Reports.