Date of Award


Degree Type


Degree Name

Doctor of Philosophy


School of Medical and Health Sciences

First Advisor

Associate Professor Peter Roberts


Despite its wide application today, in vitro fertilization (IVF) treatment continues to have relatively low efficacy, largely due to inaccuracy in selecting the best quality embryo(s) from the cohort for transfer. Novel methodologies for improved selection are being developed, and time-lapse observation of human embryos is gaining increasing popularity due to the more detailed morphokinetic information obtained plus uninterrupted culture conditions. The morphokinetic information enables the use of quantitative timings in developmental milestones of embryos and qualitative measures of abnormal biological events, to assist embryo selection/deselection. This project aimed to identify current limitations in the use of such measures and to develop recommendations for improvement in clinical application.

In the current study, most data were collected retrospectively from infertile couples seeking IVF treatment at a fertility clinic, with consent to use time-lapse incubation (Embryoscope) for embryo culture. Comparisons of time-lapse measures were made between embryos with confirmed implantation and non-implantation outcomes following uterine transfers. Thereafter, an embryo deselection model was proposed based on the retrospective findings, followed by prospective validation.

It was found in the current study that the reference starting time point (t0) in certain existing time-lapse systems was inaccurate due to (i) the early biological variations between sibling oocytes, (ii) technical limitations in current equipment and protocols, and (iii) different insemination methods used (Papers 1&2). The above variations may be minimized by using pronuclear fading (PNF, a biological time point) as t0 rather than insemination (a procedural time point) (Paper 2). An example of such application was the comparison of embryo development between patients with high and low serum progesterone levels on the trigger-day (Paper 3). Furthermore, the growth rate of embryos reported in the literature is subject to multiple clinical or laboratory factors, and this was in agreement with the present study where a published time-lapse algorithm emphasizing quantitative timing parameters was shown to lose its discriminatory power in implantation prediction when applied in two different laboratories (Paper 4). Interestingly, the qualitative measures seemed to have better inter-laboratory transferability due to the embryo growth patterns appearing independent of clinical and technical factors (Paper 4). Two novel qualitative measures were reported in the present study, namely reverse cleavage and less than 6 intercellular contact points at the end of the 4-cell stage, showing negative correlations with embryo implantation outcomes (Papers 5&6). A qualitative embryo deselection model was therefore proposed, including several qualitative measures with implantation rates being potentially increased from 22.4% to 33.6% (Paper 6). Finally, an embryo deselection model combining both qualitative and quantitative measures was reported with the use of PNF as t0, showing significant prediction of implantation outcomes in embryos regardless of insemination method (Paper 7).

In conclusion, this thesis demonstrates the usefulness of time-lapse embryo selection during IVF treatment in one specific laboratory. However, any new time-lapse parameter or model for embryo selection requires external validation by properly designed large-scale studies. Future clinical research and the development of integrated engineering and computer technology may further improve the efficacy of time-lapse selection of human embryos.


Parts of Chapters 2, 3, 4, 5 and Appendices 3-10 are not included in this version of the thesis.