Morphological Features Of The Zebrafish Embryo Image Analysis

Zebrafish has emerged as an invaluable vertebrate organism in life science research. Morphometry features in zebrafish images are important gist to judge whether the development of zebrafish embryos is health, which has important significance for research in disease modeling and drug discovery. Previous methods of zebrafish image quantification are based on human visual inspections and suffer from notable drawbacks of low accuracy, while being labor intensive and time-consuming. So, automated, computerized analysis of biology images is meaningful for researchers. In this thesis, it designs a set of reliable and effective methods to quantify curvature of zebrafish embryos tails.First, according to the characteristics of zebrafish images, we use the shape information embedded in the edge map of image, combined with morphological operations and contour following to segmente zebrafish embryos. Experiments show our segmenatation method can get correct results.Second, the principal graph algorithm is adopted to generate skeleton line of zebrafish embryos based on the idea of principal curves. To increase the robustness of our algorithm, we design a method to detect the head area whose shape appears irregular circle based on distance field so as to remove the zebrafish head skeleton line.When traditional least-squared line regression method is used to fit the tail skeleton, some exception will happen. For such case, an improvement is designed. Besides, a new line regression method based on PCA was designed.The averaged squared distance between zebrafish tail skeleton line and regressed line was used as the quantification feature to descrive the curvature of tail.Above quantification feature is used to classify zebrafish embryo images. Experimental results show our method can describe the curvature of zebrafish tail very well and is robust to different orientation of zebrafish embryos.