High throughput automated detection of axial malformations in Medaka embryo

Abstract

Fish embryo models are widely used as screening tools to assess the efficacy and /or toxicity of chemicals. This assessment involves the analysis of embryos morphological abnormalities. In this article, we propose a multi-scale pipeline to allow automated classification of fish embryos (Medaka: Oryzias latipes) based on the presence or absence of spine malformations. The proposed pipeline relies on the acquisition of fish embryo 2D images, on feature extraction based on mathematical morphology operators and on machine learning classification. After image acquisition, segmentation tools are used to detect the embryo before analysing several morphological features. An approach based on machine learning is then applied to these features to automatically classify embryos according to the presence of axial malformations. We built and validated our learning model on 1,459 images with a 10-fold crossvalidation by comparison with the gold standard of 3D observations performed under a microscope by a trained operator. Our pipeline results in correct classification in 85% of the cases included in the database. This percentage is similar to the percentage of success of a trained human operator working on 2D images. The key benefit of our approach is the low computational cost of our image analysis pipeline, which guarantees optimal throughput analysis.

Deep Learning-Enabled Morphometric Analysis for Toxicity Screening Using Zebrafish Larvae

“…So far, there have been a few studies that focused on the analysis of zebrafish phenotypes using different tools. ,,,,, One of the earliest studies used support vector machine (SVM) to extract image descriptors to recognize three basic zebrafish embryonic phenotypes, i.e., hatched, unhatched, and dead . The ability for abnormal phenotype recognition was limited due to the lack of data set of representing phenotypes.…”

“…With the continuous development of computer vision tools, Random Forrest (RF) classifiers were used to segment the contours of the swim bladder and generate spine curves to determine the developmental stages or abnormalities such as USB and spine deformation. Extremely Randomized Trees were used to differentiate individual phenotypes including normal, dead, chorion, up/down curved tail, up curved fish, short tail, hemostasis, necrosed yolk sac, and edema. ,, The DLMA workflow developed in this study demonstrated the ability to both qualitatively and quantitatively perform zebrafish larvae phenotyping. By working together with the research community to gather diverse zebrafish images, we envisage that a comprehensive model to analyze a variety of parameters in zebrafish larvae at different developmental stages could be achieved for high throughput screening in the near future.…”

“…20 Recently, studies started to look into specific organs or regions of interest to obtain quantitative measurements, such as the length and body curvature of zebrafish. 23,24 Given that exposure to toxic chemicals or pollutants would usually harvest a suite of phenotypes rather than one, it is important to develop tools that could capture the most generic phenotypes simultaneously. 25 Moreover, such analysis would enable data mining to reveal the relationship between the abnormal phenotypes and the types of chemicals, allowing for the prediction of the biological pathways involved and identification of possible toxicity mechanisms.…”

“…Machine-learning tools have shown superior capabilities to build models to differentiate morphological features in various types of organisms, including zebrafish. − Early studies focused on identifying specific embryo phenotypes such as lethality and developmental stages, but it remains challenging to handle high volume data and differentiate multiple sublethal abnormal morphologies . Recently, studies started to look into specific organs or regions of interest to obtain quantitative measurements, such as the length and body curvature of zebrafish. , Given that exposure to toxic chemicals or pollutants would usually harvest a suite of phenotypes rather than one, it is important to develop tools that could capture the most generic phenotypes simultaneously . Moreover, such analysis would enable data mining to reveal the relationship between the abnormal phenotypes and the types of chemicals, allowing for the prediction of the biological pathways involved and identification of possible toxicity mechanisms .…”

Atlas-Based Automated Detection of Swim Bladder in Medaka Embryo

“…We also observe a high specificity of 96% and an acceptable sensitivity of 90%, which suggests that the use of this method is accurate for the global screening test that is to be developed. Indeed, the swim bladder detection assay is intended to be a part of a series of morphological abnormalities detection assays whose each specificity needs to be maximized in order to maximize the accuracy of the whole detection assay and not eliminate a too important number of chemicals [11,27]. The sensitivity of each individual test is expected to improve the sensitivity of the whole test.…”