AI Embryo Grading in IVF: What a Laboratory Score Can and Cannot Predict

Some systems analyse visible morphology, developmental timing or image features that have been labelled in training data. Their output may be a grade, score or rank indicating which embryos appear more promising under that model. Different laboratories use different culture systems, cameras, clinical protocols and software versions. A score from one vendor should not automatically be compared with another vendor’s score. The training population and the outcome used to build the model matter. A ranking tool cannot directly observe every cause of implantation failure or pregnancy loss. It also cannot tell a patient whether a particular embryo would lead to a healthy baby.

AI may help an embryologist review images consistently or spend less time on routine scoring. Those operational advantages can be valuable to a laboratory without necessarily improving the patient’s chance of a live birth. The American Society for Reproductive Medicine has described promise for workflow and grading consistency while noting that comparative trial evidence has not reliably shown improved outcomes over conventional embryo assessment. Ask whether the clinic is describing fewer staff minutes, an accuracy metric or an actual patient outcome. A promotional claim of “better embryo selection” should state its comparison group and measured endpoint. A statistical association between score and outcome does not establish that using the score changes clinical outcomes.

PGT-A involves testing biopsied embryonic cells for certain chromosomal findings; PGT-M concerns a specific known monogenic condition when appropriate. AI image ranking generally examines appearance or development patterns and does not replace a clinically indicated genetic evaluation. Neither AI scores nor embryo grades can guarantee a normal chromosome result or a birth. Genetic testing itself has limitations and should not be offered universally merely because software labels an embryo as “uncertain.” Ask the doctor which question each proposed test answers and whether that answer would alter a decision. Be alert to bundled fees that make several unrelated add-ons sound like one indispensable technology.

Ask for the exact system name, its role in embryo selection, the validation evidence, who can override a rank and how disagreement is documented. What information about embryos or patients is stored, processed or shared with the software provider? How are privacy and data retention handled? Some software may not be separately billed, while other clinics charge; obtain written terms rather than inventing a market price. The embryologist should explain whether embryo culture conditions, grading criteria and laboratory quality controls are standard independently of the algorithm. A patient can decline an optional, unproven add-on without assuming the clinical team cannot provide conventional fertility care.

HomeIVF is a doctor-led fertility care coordination model intended to reduce avoidable travel for suitable consultations and monitoring steps where available. A remote discussion can help patients understand an embryology report or seek a second opinion, but it does not perform embryo imaging or grading at home. Egg retrieval, fertilisation, embryo culture, possible biopsy and transfer remain in appropriately equipped facilities. Do not infer that HomeIVF owns or uses any particular AI embryo-ranking product; that would require separate verification. The most valuable outcome of a discussion may be a clearer treatment choice, not a technology purchase. Ask what happens if software and the clinical embryologist rank embryos differently.

If a clinic says its AI system improves embryo selection, ask whether “improves” refers to grading agreement, laboratory efficiency, clinical pregnancy, ongoing pregnancy or live birth. Those are not identical outcomes. Then ask whether the comparison came from a randomized trial, a retrospective dataset or a vendor presentation. Even a statistically strong prediction within one laboratory may behave differently elsewhere. Find out who reviews cases where the tool and embryologist disagree, whether the algorithm has changed versions and whether patients can opt out of an optional scoring service. A written evidence summary is more useful than a 99-point scale with no defined endpoint. Also ask whether the same conventional embryo evaluation and standard-quality lab culture would be available without paying extra for the software.

• Ask which algorithm is used and what it measures.
• Ask the clinician whether the software result changes a recommended decision.
• Request evidence from appropriate comparative clinical studies.
• Verify the authorised laboratory and embryologist responsible.
• Check consent, data handling and optional charges before agreeing.