Medical imaging and visualisation

The assessment of embryo morphology may be more accurately achieved using medical imaging techniques such as segmentation alongside conventional grading systems. Segmentation is a method utilised for the visualisation of 3-dimensional images of various tissues including brain, cardiac and tumour tissues as well as embryos, and allows images to be partitioned into regions to enable the extraction of specific features and obtain informative measurements (Rogowska 2000). One such application of medical imaging assessed the thickness of the ZP of human embryos, and involved edge mapping of the externa and interna membranes with an automated active contour model (snake segmentation) to distinguish the boundaries, and allow the exact measurements to be obtained. The ZP thickness variation was then retrospectively compared with pregnancy results in order to identify a correlation, and provides evidence for utilising this technique to provide additional selection criteria for embryos prior to uterine transfer (Moraleset al

2008).

There are several different classifications of segmentation techniques depending upon their intended application including but not limited to, manual, automatic and semi-automatic, pixel-based and region-based, manual delineation, and classical (Rogowska 2000). Edge-based segmentation identifies distinct boundaries between regions with different characteristics, and this technique is suitable for imaging embryos since they contain several membranes enclosing various cell structures , for example the ZP surrounding the cytoplasm of zygotes, individual blastomere

membranes in cleavage stage embryos, and the cells of the inner cell mass and trophectoderm of blastocysts. Definition of boundaries depends upon the local pixel intensity gradient, with images created according to both the magnitude and

direction of the gradient, and by calculation of the summation of localised pixel intensities (Rogowska 2000). Semiautomatic edge linking, where the operator manually draws the edge when the automatic tracing becomes ambiguous, circumvents the problem of incomplete enclosure of the cell often seen with segmentation. Multimodal or multispectral techniques incorporate data from a series of images during which the intensity gradients alter but the anatomical structure remains constant, such as those acquired in embryo time-lapse recordings. These allow the incorporation of data from a spectrum of features, providing an array of information for use in segmentation and imaging. The

parametric analysis technique, where the pixel intensity for each region of interest is plotted against time, is commonly employed when analysing serial images procured over extended time periods. Parameters are selected according to the functional characteristics of the object in question (eg an individual blastomere), such as time of occurrence of minimum and maximum intensities, and an image calculated for each of the parameters, hence the term parametric imaging. Edge detection segmentation can be used to create 3-dimensional images, and combined with 3-dimensional rendering, allows for detailed analysis of image structures (Rogowska 2000).

Visualisation aims to develop tools that allow the “inside” of complex living systems to be observed and explored, and involves the use of computing environments,

graphics hardware and software that facilitate the interaction between humans, machines and data. Visualisation techniques enable the generation of realistic 3- dimensional images, development of automated methods for manipulation of multidimensional images and their associated parametric data, advancement of measurement tools for analysis of images, and the design and validation of models that improve the interpretation of diagnostic and clinical applications (Robb 2000). Virtual reality facilitates interactive visualisation, allowing the operator to control detailed computer-generated 3-dimensional images that mimic real-life situations, enabling manipulation of images in such a way that the operator may “enter” the visualisation and change their viewpoint as if the image were a real object (Robb 2000). This has huge implications in the development of tools for the purposes of teaching and demonstrating, as well as for diagnostic applications, and virtual procedures have been found to provide accurate, reproducible visualisations that are clinically useful and minimally invasive (Robb 2000). This suggests that the inclusion of virtual models in the routine clinical environment is a very real possibility, and may prove invaluable in the future of medical advancement. Visualisation in medicine may be characterised, according to the complexity of the resulting user interaction, as illustrative, investigative or imitative. Illustrative visualisation presents extracted data in the form of three-dimensional displays, whilst investigative visualisation is more focussed on explorative techniques such as multidimensional interpretations useful for clinical tools. Imitative visualisation refers to virtual reality systems, simulators and modelling (Solaiyappan 2000).

The use of computer software for the development of advanced image processing systems has led to an increase in the application of digital imaging to human medicine. Medical visualisation describes the generation of three-dimensional (3D) models using digital algorithms, based upon living specimens, to increase the understanding of biological structures and their functions (Klauschenet al2009), and is a useful tool when attempting to understand the mechanisms of

development and/or abberant behaviours. Therefore, development of semi- automated human embryo analysis tools may assist in generating high volumes of data to increase understanding of cell division patterns in relation to clinical outcomes, and may improve our ability to select the most viable embryos from a cohort from an early stage of development.

Aims

The overall aim of this project was to investigate possible molecular, morphological and kinetic markers of preimplantation human embryo viability, with a view to increasing our understanding of human embryo development and improving embryo selection methods in the clinical embryology laboratory.

Briefly, the main aims of this project were:

• To detect and confirm the distribution of the actin cytoskeleton in cleavage stage, compacting, and arrested human preimplantation embryos and blastocysts.

• To establish whether the regulatory proteins STAT3 and leptin are polarised within human oocytes and preimplantation embryos

• To determine the number of cells in the inner cell mass and the

trophectoderm in human blastocysts duringin vitrodevelopment, and to consider morphological characteristics of blastocysts and their relationship with quality and possible clinical outcomes.

• To investigate the occurrence of chromatin-containing fragments and micronuclei in human preimplantation embryos of all developmental stages

• To establish whether STAT3 activation may represent a mechanism by which the embryo and endometrium communicate at implantation in humans.

• To detect the presence of leptin receptor mRNA on human blastocysts in order to determine whether the leptin system plays a role in the activation of STAT3 in human pre-implantation embryos.

• To investigate the interaction between embryos and Ishikawa cells, as a model of the endometrial epithelium.

• To identify potential markers of embryo viability and predictors of embryo development by assessment of human preimplantation embryos using time lapse monitoring.

• To assess the potential effect of deliberately damaging the daughters of the first dividing cell at the third cleavage division (4-8 cells) with respect to ensuing cleavage patterns and blastocyst formation, to determine whether damage of specific blastomeres may impact on axis establishment and subsequent embryo development.

• To identify the initial site of blastocyst hatching from the ZP with respect to the position of the inner call mass, to establish whether there is any association between the site of hatching and pregnancy potential.

• To explore different options for computer-assisted image analysis of human embryos at different stages of development and using different image types.

• To apply the most effective semi automated computer-assisted options to investigate the nuclear to cell ratio in cleavage stage human preimplantation embryos, and the nuclear volumes of different cell lineages in blastocysts, using data derived from a semi-automated computer programme developed in

collaboration with colleagues at the Warwick Digital Laboratory, and Manchester University School of Computing.

Chapter Two