The fourth-day question
Couples going through IVF face a series of suspenseful waits. One of the hardest comes around Day 4 after fertilization. By then, the lab has watched embryos grow for nearly a week. Some look promising. Others look uncertain.
The team has to decide which embryo to transfer first. Pick the right one and the chances of pregnancy go up. Pick poorly and you may use the strongest embryo too late.
A new study suggests that two simple features may help make that call.
Even with modern IVF, only about 30 to 40 percent of cycles result in a live birth. A big part of that gap comes down to embryo selection. Labs use a mix of visual grading and time-lapse imaging to judge which embryos are most likely to develop further. But there's no perfect rule, and clinics still rely on a lot of judgment.
For patients, that means uncertainty at the most emotional point in the journey. For doctors, it means the same.
A more reliable Day 4 score could help both sides plan.
The old way versus the new way
Until now, deciding which Day 4 embryos were most likely to keep developing relied mostly on how they looked under a microscope and on a lab's experience.
The new model takes a different approach. It uses just two pieces of information: how many cells the embryo had on Day 3, and a computer-vision score called iDAScore that rates embryo quality automatically. Combine the two, and the result predicts whether that embryo is likely to reach the early blastocyst stage by Day 4 — a key milestone on the road to a successful pregnancy.
Think of growing an embryo like baking a delicate dough. Some doughs rise quickly and evenly. Others stall and need more time, sometimes ending up just as good and sometimes not.
The number of cells on Day 3 tells you how fast the dough is rising. The iDAScore is like a baker checking that the rise is also healthy and not lopsided. Neither alone tells the full story. Together, they paint a clearer picture.
A higher cell count on Day 3 (more than ten) and a high iDAScore (above 4.8) each independently raised the odds of reaching the early blastocyst stage. Together, they were the strongest predictors.
The study snapshot
The team studied 2,557 patients scheduled for fresh Day 4 embryo transfers at a single Chinese hospital between January 2021 and December 2025. They split the data into two groups — about 1,800 patients to build the model and 770 to test it. About 44 percent of embryos in both groups went on to form an early blastocyst by Day 4.
The two-factor model held up well. In the building group, it correctly distinguished embryos that would reach early blastocyst stage from those that wouldn't about 78 percent of the time. In the validation group, that number was a similar 77 percent. Both numbers are considered solid for a clinical prediction tool.
In practical terms, the model tended to flag the right embryos as most likely to develop, with a sensitivity around 65 percent and a specificity around 80 percent. That balance is useful in a clinic setting because it limits the number of embryos that get unfairly written off.
This isn't a guarantee that any single embryo will lead to a pregnancy.
Where this fits in the bigger picture
Embryo selection is one of the fastest-moving areas of fertility medicine. AI tools now analyze time-lapse video of every embryo and produce automatic quality scores. Genetic testing can flag embryos with chromosomal problems before transfer.
This new model fits between those tools. It's simple, doesn't require advanced testing or expensive software beyond what many clinics already use, and gives a single number to help time the transfer. That makes it more accessible than some of the higher-tech options, especially in clinics with tighter budgets.
If you're going through IVF, this study won't change your treatment overnight. But it adds to the evidence supporting Day 3 cell count and AI-based scoring as legitimate tools to time embryo transfers and pick which embryo to use first.
If your clinic uses time-lapse imaging or any AI scoring system, it's worth asking how those scores factor into the choice of which embryo to transfer and when. Decisions about Day 3 versus Day 5 transfers may benefit from this kind of model.
The data come from one fertility clinic in China. Practices, patient populations, and lab equipment vary widely between clinics, so the model may need to be retrained on local data before it can be trusted elsewhere. The team also tracked whether embryos reached the early blastocyst stage, not whether they led to live births. Those are related but not the same. A larger, multi-center follow-up will be needed to see if the model improves real pregnancy outcomes.
The researchers are now planning to test the model in multiple clinics and to extend it from predicting blastocyst formation to predicting pregnancy. If those efforts succeed, IVF teams could have a quick, low-cost way to make Day 4 decisions more confidently. The next year or two should bring more data.