Does conventional morphological evaluation still play a role in predicting blastocyst formation?

Area of Science:

• Reproductive medicine
• Embryology
• In vitro fertilization (IVF)

Background:

• Advanced technologies like AI and time-lapse imaging are emerging for blastocyst formation prediction.
• Conventional morphological evaluation remains a widely used standard in IVF clinics.
• This study assesses the predictive capability of traditional embryo morphology assessment for blastocyst development.

Purpose of the Study:

• To evaluate the predictive power of conventional embryo morphological assessment for blastocyst formation.
• To compare the efficacy of morphology-based models with those incorporating additional patient and cycle data.

Main Methods:

• Retrospective analysis of 15,613 IVF cycles undergoing blastocyst culture (2013-2020).
• Generalized estimating equations (GEE) model developed using morphological parameters (cell number, fragmentation, symmetry, early cleavage).
• Comparative models built using LASSO regression and XGBoost incorporating 27 additional features (patient characteristics, cycle parameters).

Main Results:

• The morphology-based GEE model predicted blastocyst formation with an AUC of 0.779 and accuracy of 74.7% in the validation set.
• Models combining morphology with patient/cycle data (LASSO, XGBoost) showed similar predictive power (AUCs 0.78 and 0.754).
• Morphological parameters alone strongly correlated with observed blastocyst yield (r=0.897) and predicted transfer cancellation (AUC=0.926).

Conclusions:

• Routine embryo morphology assessment is a feasible and valuable tool for guiding embryo transfer decisions.
• Integrating patient characteristics and cycle parameters into predictive models did not significantly enhance accuracy over morphology alone.