A study of 280 fresh blastocyst transfers found faster expansion in embryos that led to pregnancy.
Time-lapse imaging and artificial intelligence may add information to conventional embryo grading, but prediction remains limited.
Which embryo should be transferred to maximize the chance of pregnancy after IVF? Embryologists assess appearance, developmental stage and cellular characteristics. A study suggests that how quickly an embryo expands during culture may also be associated with pregnancy outcomes.
Researchers at Erasmus University Medical Center in the Netherlands analyzed time-lapse embryo images using artificial intelligence. Blastocysts that led to pregnancy expanded significantly faster than those that did not.
Published on October 21 in Reproductive Biology and Endocrinology, the study explored whether measuring growth over time could complement assessment of an embryo’s appearance.
The same grade can hide different growth patterns
Embryo quality is important when assessing the likelihood of pregnancy after IVF. Blastocysts, usually reached around five to six days after fertilization, are graded using features such as their inner cell mass, trophectoderm and degree of cavity expansion.
The widely used Gardner grading system assesses expansion and cellular appearance.
However, assessments made at a particular time cannot capture every aspect of development. Embryos with the same grade may follow different growth trajectories.
As a blastocyst develops, fluid accumulates inside it and its cavity expands. Its size and shape consequently change over time.
The researchers investigated whether measuring the rate of expansion, rather than size at a single moment alone, could provide additional information.
Faster expansion was associated with pregnancy in 280 transfers
The team retrospectively analyzed data from 280 patients who underwent a single fresh blastocyst transfer after time-lapse culture.
Time-lapse incubators take images at regular intervals without removing embryos from the incubator. This allows their development, from cell division to blastocyst formation, to be observed over time.
The researchers applied deep learning to the images, automatically identifying blastocyst boundaries and measuring their cross-sectional area.
The analysis covered the interval from the start of blastocyst formation to approximately 112–116 hours after fertilization. Changes in area were used to construct expansion curves and extract growth-related features.
Blastocysts that led to biochemical pregnancy and those that led to clinical pregnancy with a confirmed fetal heartbeat both expanded significantly faster than blastocysts that did not lead to pregnancy.
Expansion rate may therefore provide one developmental marker associated with pregnancy outcomes. This observational study does not establish that faster expansion itself causes pregnancy success.
Accurate image measurement is not the same as accurate pregnancy prediction
A notable feature of the study was its numerical measurement of embryo growth.
Conventional assessment relies on embryologists examining images and morphology and applying their experience. Here, artificial intelligence automatically measured cross-sectional area and calculated changes over time.
The segmentation model achieved a Jaccard index of 97.6%, a measure of how accurately it identified the blastocyst region in an image.
Accurately measuring an embryo’s size and predicting whether it will lead to pregnancy are different tasks.
Models combining patients’ clinical information with blastocyst expansion features achieved ROC-AUC values of 0.69 for biochemical pregnancy and 0.67 for clinical pregnancy with a fetal heartbeat.
ROC-AUC measures how well a model distinguishes outcomes: 0.5 corresponds to chance-level discrimination, while values closer to 1.0 indicate better discrimination.
An AUC of 0.67 suggests that expansion features provide some predictive information, but remain insufficient to determine pregnancy success reliably on their own.
The researchers also called for further validation in larger patient groups.
From a snapshot of shape to the process of growth
The study illustrates one possible direction for embryo assessment in IVF.
Assessments have traditionally emphasized morphology and developmental stage at particular moments. Combining time-lapse technology with artificial intelligence makes it possible to analyze the developmental process itself.
Expansion rate is particularly interesting because it can be measured from images without an additional invasive test on the embryo.
However, faster expansion does not necessarily mean that an embryo has normal chromosomes. This study did not establish how accurately expansion predicts chromosomal status or live birth.
The analysis was also limited to 280 patients receiving a single fresh blastocyst. Further research is needed to determine whether the findings apply to frozen transfers or other patient populations.
Experienced embryologists remain important in embryo selection. Morphology and developmental status must be considered alongside factors such as patient age and treatment history.
The contribution of this research is not proof that artificial intelligence can replace embryologists’ judgment. It is the provision of objective measurements of developmental speed that could broaden the information available for selection.
When assessing embryo quality, the way an embryo grows may matter alongside how large it becomes. The study adds evidence for investigating laboratory assessment that measures development over time as well as appearance.
Source: This article is based on “Quantitative blastocyst expansion features derived via deep learning from time-lapse sequences predict clinical outcomes,” by researchers at Erasmus University Medical Center, published in Reproductive Biology and Endocrinology on October 21, 2026. DOI: 10.1186/s12958-026-01615-z.
The illustration was created using generative AI (ChatGPT, OpenAI). Any people depicted are fictional.
Related reading on assessing evidence before adopting additional IVF techniques: what a review of IVF add-ons found about live birth.
Copyright © 2026 Fertility Information TokTok Fertility News | THE FERTILITY NOTES. All rights reserved. Unauthorized reproduction, copying, redistribution, scraping, AI training, and other data use of any articles, photographs, images, or other content are strictly prohibited. Violations may result in legal liability.