Embryo classification beyond pregnancy: early prediction of first trimester miscarriage using machine learning

Area of Science:

• Reproductive Medicine
• Embryology
• Artificial Intelligence in Healthcare

Background:

• First trimester miscarriage is a significant concern in in vitro fertilization (IVF) treatments.
• It affects a substantial proportion of clinical and recognized pregnancies following IVF-ET.

Purpose of the Study:

• To develop a machine learning classifier for predicting the risk of first trimester miscarriage.
• The classifier utilizes time-lapse imaging of preimplantation embryo development.

Main Methods:

• Retrospective analysis of 391 women undergoing ICSI and fresh embryo transfers.
• Embryo morphodynamic features, including pronuclei dynamics, were analyzed.
• XGBoost and random forest models were trained and validated using Monte Carlo cross-validation.

Main Results:

• A subset of six non-redundant morphodynamic features accurately predicted miscarriage risk.
• Features related to nucleolus precursor bodies and pronuclei dynamics were highly predictive.
• The developed models achieved an AUC of 0.68-0.69 for miscarriage prediction.

Conclusions:

• A decision-support tool was created to identify embryos at high risk of miscarriage.
• Prioritizing embryo transfer based on predicted miscarriage risk can enhance live-birth rates.
• This approach is expected to shorten the time to pregnancy for infertile couples.