Mechanics of Human Pre Implantation Development
- Conditions
- Infertility
- Registration Number
- NCT06408389
- Lead Sponsor
- Assistance Publique - Hôpitaux de Paris
- Brief Summary
The time-lapse is a closed tri-gas incubator of the latest generation that provides optimal and stable culture conditions for the culture of embryos in In Vitro Fertilization (IVF). The integration of a camera within this incubator allows for continuous image capture, thus facilitating the monitoring of the entire embryonic development, from the day of fertilization to the moment of transfer into the uterus.
The contribution of the time-lapse system allows an evaluation of the embryos not only by their morphology, but also by their cell division kinetics, both being direct markers of cell mechanics. Together, these morpho-kinetic data finally allow for the best identification of embryos with greater implantation potential. Time-lapse imaging represents a further step towards an objective assessment of the embryo, but inter- and intra-embryologist variations in annotations partly compromise this objectivity. In addition, many decision algorithms based on the evaluation of morpho-kinetic parameters have been developed, but the lack of reproducibility from one Assisted Reproductive Technology (ART) center to another is a hindrance to the generalization of any particular algorithm. The aim of this retrospective study is to determine morpho-kinetic factors predictive of implantation using machine learning and to link these factors to human embryo mechanistic properties.
- Detailed Description
The time-lapse is a closed tri-gas incubator of the latest generation that provides optimal and stable culture conditions for the culture of embryos in In Vitro Fertilization (IVF). The integration of a camera within this incubator allows for continuous image capture, thus facilitating the monitoring of the entire embryonic development, from the day of fertilization to the moment of transfer into the uterus.
The contribution of the time-lapse system allows an evaluation of the embryos not only by their morphology, but also by their cell division kinetics, both being direct markers of cell mechanics. Together, these morpho-kinetic data finally allow for the best identification of embryos with greater implantation potential. Time-lapse imaging represents a further step towards an objective assessment of the embryo, but inter- and intra-embryologist variations in annotations partly compromise this objectivity. In addition, many decision algorithms based on the evaluation of morpho-kinetic parameters have been developed, but the lack of reproducibility from one Assisted Reproductive Technology (ART) center to another is a hindrance to the generalization of any particular algorithm.
Machine learning is one of the main methods of data analysis that could define algorithms that are unbiased, more robust and applicable to all centers. But the optimal algorithm is not yet defined. Recently, an artificial intelligence approach applied to a large collection of time-lapse embryo images was developed to determine the embryo with the highest grade of evolution, with an AUC\> 0.98. Using clinical data, the authors created a decision tree to integrate embryo quality and female age and identify the chances of pregnancy. However, this approach did not take into account the whole kinetics of development, focusing on certain particular stages, nor the influence of parental and extrinsic factors other than age.
The aim of this retrospective study is to determine morpho-kinetic factors predictive of implantation and embryo development in IVF/ICSI using machine learning algorithms and relate these morpho-kinetic factors to the mechanical characteristics of cells.
Recruitment & Eligibility
- Status
- NOT_YET_RECRUITING
- Sex
- All
- Target Recruitment
- 1500
- Couples enrolled in an IVF process and with embryos cultured in time-lapse
- Couples informed non opposed to research
- Couples opposed to research
- Couples under curator or tutorship
- Couples under state xxx
Study & Design
- Study Type
- OBSERVATIONAL
- Study Design
- Not specified
- Primary Outcome Measures
Name Time Method Prediction of embryo implantation by the machine learning algorithm from embryo morphokinetic parameters and patient health data. 2 years The algorithm outcome will be evaluated retrospectively on human embryos which have been transferred, for which we know whether its implantation was successful and led to birth. Using embryo morphokinetic and health patient data, we will predict a probability of implantion and compare its value (\<0.5: no implantation or \>0.5: implantation) to the true result of the embryo transfer (no implantation or implantation). This will allow us to evaluate the potential of the algorithm to support clinical decision-making in the future.
- Secondary Outcome Measures
Name Time Method Link between couples' parameters and embryo morpho-kinetics 3 years Identify statistical links between parental or extrinsic factors (linked to the attempt) and morpho-kinetic parameters :
Male : cause of male infertility (oligoasthenoteratospermia (yes/no))Software developing 3 years Develop decision support software for the benefit of embryologists as part of an ART attempt
Statistical correlation between embryo implantation prediction and morphokinetic parameters 3 years The machine learning algorithm (neural network) trained for the primary outcome will be used to quantify the average percentage contributed by each morphokinetic parameter in the prediction of implantation success or failure ("explainable artificial intelligence"). This will allow us to determine what morphokinetic parameters influence the most positively or negatively the prediction of implantation.
Embryo reconstruction in 3D 3 years Develop an automatic method for reconstructing the morphology of embryos in 3 dimensions from 2-dimensional images in transmitted light (Geri incubator or Embryoscope type)