Review of Bioinformatics and Biometrics (RBB) Volume 1 Issue 1, December 2012 www.seipub.org/rbb 1 Use of Concave Corners in the Segmentation of Embryological Datasets Kieran Rafferty 1 , Sarah Drury 2 , Geraldine Hartshorne 2 , Silvester Czanner *1 *1 School of Computing, Mathematics and Digital Technology, Manchester Metropolitan University, Manchester , M1 5GD, UK 2 Division of Reproductive Health, Clinical Sciences Research Laboratories, Warwick Medical School, University of Warwick, Coventry, CV2 2DX, UK 1 Kieran.Rafferty@stu.mmu.ac.uk; 2 S.L.Drury@warwick.ac.uk; 2 Geraldine.Hartshorne@warwick.ac.uk; *1 S.Czanner@mmu.ac.uk Abstract Embryologists require identifiable factors when analysing embryos grown in‐vitro that can better predict an embryos successful development in the uterus. It has been proposed that the volume of various embryo features, such as blastomeres, within a developing of embryo could be used as an indicator of viability. An initial interface has been developed using existing image processing tools to segment z‐stacks produced via confocal microscopy and create a set of volumes for analysis. A further addition to these tools was developed with the intent of improving the accuracy of the segmentation process; specifically, in identifying individual blastomeres whose proximity to neighbouring blastomeres resulted in the false identification of a single segment. A method has been proposed that uses a contour produced by an earlier segmentation procedure and identifies concave regions along the contour. These are interpreted as probable indicators of the location of boundaries between different blastomeres. By using points along the contour in these concave regions we attempt to extrapolate the boundaries between blastomeres. Keywords Embryological datasets; Image segmentation; Embryology; Blastomeres; Concave corners Introduction Assisted reproductive technology (ART) is a term that covers a range of methods used primarily in infertility treatments achieving pregnancy through artificial or partially artificial means. Methods of ART include fertility medication, which is used to stimulate follicle development of the ovary and in vitro fertilisation (IVF), where an egg is fertilised outside of the body in a laboratory setting in vitro and then transferred to the patients uterus. The emphasis of current research in the field of IVF is to determine the qualities by which an embryo can be evaluated that would indicate it is likely to result in a successful pregnancy. (Filho, 2010) refers to a study that compared the grades compiled from a number of practicing embryologists against a control. Substantial differences between morphological scores were observed by as much as two grades, despite using the same grading system. The limited time between insemination and implantation is a large determining factor in the ability of embryologists to make quantified assessments of developing embryos. Embryos at the blastocyst stage may have greater than a hundred constituent cells and accurately measuring the size of each would be a long and tedious process. The development of computer algorithms to take these measurements would be ideal in eliminating this hurdle. Using image processing it is hoped that potential indicators can be quantified and accurately measured. Grading of the blastocyst has become more popular in recent years (Filho, 2010). The blastocyst stage of embryo development is usually reached around five days after insemination and some nonviable embryos may undergo apoptosis, automatically eliminating themselves from the embryo selection process. By waiting for embryos to reach the blastocyst stage of development embryologists have a better idea of the potential viability of an embryo. Despite this, embryologists still require further identifiable factors that can better predict an embryos successful development in the uterus. A large number of studies that have attempted to automatically segment embryos have used images produced from HMC optical to do so (Morales et al 2008, Pederson et al 2003, Giusti et al 2010, Karlsson et al 2004). Fewer have attempted to use confocal imaging though further examples of confocal microscopy segmentation are found outside of