AI embryo grading explained: from Day 1 to the Day-5 blastocyst

Choosing which embryo to transfer is one of the most consequential decisions in an IVF cycle. Traditionally, that choice rests on an embryologist's visual assessment under the microscope — expert work, but inherently subjective. Studies have long shown that agreement between experienced embryologists grading the same embryo is often only 50–70%. AI-assisted grading exists to make that decision more objective, consistent and explainable.

What "grading" actually measures

An embryo isn't judged on a single snapshot. Its quality is a story that unfolds over five days, and modern grading looks at that whole trajectory:

• Day 1 — fertilisation. Confirming normal fertilisation (two pronuclei) and early morphology.
• Days 2–3 — cleavage. The number and symmetry of cells, and the degree of fragmentation.
• Day 4 — morula. Compaction, as cells begin to bind tightly together.
• Day 5 — blastocyst. Expansion of the fluid-filled cavity, the inner cell mass (which becomes the baby) and the trophectoderm (which becomes the placenta).

Each stage carries signal about developmental competence — and each is a point where subtle, easy-to-miss features matter.

How AI reads the embryo

Garbha's EmbryoScore uses a multi-stage convolutional neural network trained on a validated dataset of more than 36,000 embryo images. Rather than a single verdict, it analyses the developmental sequence:

• Morphokinetic annotation — tracking when key events happen, not just how the embryo looks at one moment.
• Fragmentation index — quantifying cellular fragmentation objectively instead of by eye.
• Developmental scoring — combining these signals into a consistent grade, with 93% embryo-selection accuracy.

Because the model applies the same criteria every time, it removes the shift-to-shift and site-to-site variability that manual grading can introduce.

Why explainability matters

A grade you can't interrogate is hard to trust in a clinical setting. Garbha surfaces Grad-CAM heatmaps that highlight the regions of the embryo driving each decision, alongside the live annotation. Embryologists can see the reasoning — which makes the tool useful for training and quality review, and auditable for regulators. It complements the embryologist's judgement rather than replacing it.

The bottom line

AI embryo grading doesn't change what makes a good embryo — it changes how reliably and transparently we can identify one. By reading the full Day 1-to-Day 5 journey with consistent, explainable criteria, it helps labs prioritise the most viable embryo, supporting better implantation and live-birth rates.

Further reading

• World Health Organization — Infertility
• ESHRE — European Society of Human Reproduction and Embryology
• ASRM — American Society for Reproductive Medicine

Want to see explainable embryo grading on your own cases? Explore the technology or book a demo.