Revolutionising Reproduction: The Role of AI in IVF

Success rates are notoriously low in IVF outcomes and innovation in the field of IVF is disappointingly slow. Are we finally on the precipice of better-quality treatment and improved IVF outcomes? How is AI used in IVF clinics? 

The most recent figures from the HFEA tell us the current live birth rate per embryo transferred stands at 25% and 19% respectively for patients aged 35-37 and 38-39. IVF as a science has been stagnant and the consideration of AI in an IVF clinic setting is long overdue. Globally, current IVF success rates sit at ~30% and with so much ongoing research to improve these rates it is no surprise that the benefits of AI and machine learning are now being considered in the IVF clinic.

One in six couples will struggle with infertility and IVF is one of several treatments available to help bring a baby home. IVF is the process of removing an egg from the ovary of a woman and fertilising that egg with sperm in a laboratory setting. The resulting embryo is then transferred back into the uterus of the woman, either as a fresh embryo transfer or a frozen one. So why doesn’t it always work, why are success rates low and why is there such variability of success rates amongst clinics? These rates are contingent on both the cause of infertility and the age of the woman undergoing treatment, but there’s also the subjectivity of care, embryo selection and treatment protocols. Can AI help reduce this variability and help physicians increase IVF success rates? 

AI is the term to describe mathematical algorithms that try to automate decisions or analyses performed by a clinician or embryologist. The vast amount of data an algorithm can process and categorise means it could have great use in an IVF by factoring in data from other IVF cycles and using that to suggest a personalised IVF protocol, or by selecting the best embryo for transfer. These are currently the best two use cases for AI in the IVF clinic. 

It is thought that the current approach, with humans making the decisions, means their subjectivity factors into decision-making and that is why there are such degrees of variation between clinics.  This is where AI steps in as it would remove the subjectivity of human assessment from the decision-making, and objectively rank embryos or determine patient protocols, based on data.

Embryo Selection

Embryo selection is the area of AI that has been given the most consideration and it is likely this is how we’ll first see AI most commonly used in clinics. 

AI vs the Human Eye

Your doctor has a complex set of clinical decisions to make to deliver optimal care and ensure your best chances of not only conceiving but bringing a baby home. Selecting the most viable embryo to help this happen is a crucial job for the embryologist, done manually by observing embryos through a standard microscope-mounted camera or a time-lapse incubator system. Based on appearance alone and chromosomal testing results, where available, the best embryo is selected for transfer.

This manual and time-consuming process is susceptible to bias and error. What do we mean by that exactly?

How and where an embryologist was trained, the standard operating procedures of the clinic they work in, how that clinic grades and selects embryos based on morphological features - these are all variable factors that can impact how the embryo is selected and it is this inherent bias that the use of AI hopes to remove.

There’s also a limited amount of data an embryology team in one clinic can draw upon to influence their decision-making:

“The amount of data about embryos, past patients, and successful live births available to any single doctor is very small, so it’s hard for them to generalize about what indicates that a fertilised egg is viable.”

David Silver, founder of Embryonics

Fed with enough data, AI would use pattern recognition and reference data sets of potentially hundreds of thousands, even millions, of IVF cycles to recommend which embryos could be the most successful for a particular patient. The human eye is simply not capable of assessing embryo viability to the degree that an AI model could, these are complexities that humans just cannot compute.

There’s also the cost element of this part of your IVF treatment, it is time-consuming for your embryologist to monitor and assess your embryos. The average fee per cycle associated with this run to around £5,000 and, in the long term, AI should be able to reduce those costs.

The potential of AI to prioritise the most viable and transferable embryos is an exciting prospect and it’s the kind of personalisation in IVF treatment we want to see more of. Picking an embryo with the best chance of survival doesn’t correspond to bringing a baby home, but it’s a pretty great place to start. 

IVF Treatment Protocols

Protocols vary widely and it can be a case of ‘trial and error’ before a more personalised protocol is achieved, with the average patient having three to four IVF cycles. The financial and emotional impact of multiple rounds is obvious, and the lack of a more personalised approach is an often-quoted criticism of the IVF sector. 

The human bias element in preparing patient protocols stems from physicians’ own clinical experience and standard in-house clinic practices and training. AI could help physicians in formulating an optimal, personalised fertility treatment plan based on patient characteristics (age, egg quality, medical history etc.), again by benefiting from large amounts of cycle data held in an AI model that they otherwise would not have access to.

Adding an AI tool into the mix has the potential to remove that subjectivity. It can even review similarities among patients and prepare simulations to advise what would happen if another protocol was chosen. It could also assess data from patient records to better understand factors predictive of live birth and pregnancy. 

Is AI a Form of IVF?

Not quite, although maybe it has the potential to be. LifeWhisperer showed their AI tool resulted in a 12% in time-to-pregnancy for those undergoing IVF. Another AI tool outperformed 15 embryologists in predicting embryos with the greatest implantation potential. There are a great number of new AI tools launching in the IVF sector (almost too many to list) and so far AI has been developed successfully to predict blastocyst formation and fertilisation potential. 

Is My IVF Clinic Using AI? 

Whilst there are a great number of AI tools available, the use of AI in the clinical setting is not yet mainstream, but things will move fast. Here’s what we likely need to see before we can expect mainstream adoption by IVF clinics:

We Need More Clinical Evidence

The IVF sector expects to see high-quality randomised controlled trials (RCTs) to show efficacy and improvement in fertility outcomes before using a new tool such as AI. The problem with this approach when it comes to AI? By the time the RCT is published, your AI algorithm is already well out of date. There’s also a wider debate as to whether RCTs by design are suited to studying clinical decision support tools such as these, so we might need to accept that RCTs cannot be the only method of validating this new technology.  


The independent regulator of fertility treatment in the UK, the Human Fertilisation and Embryology Authority (HFEA) uses a red/amber/green system to assess any new treatments or treatments in addition to routine fertility treatment. These are referred to as “add-ons” and before they’d consider giving AI tools for embryo selection the green light, they require this to be tested in RCTs. There is also a range of autonomy in AI technologies, all with different levels of risk, but so far it seems they’re all being categorised together for regulatory purposes. The new sandbox approach proposed by the HFEA in their latest consultation to update UK fertility laws could be ideal for AI, this would permit the regulator to approve the use of various AI for a specified period before assessing real-world evidence and should allow for faster-paced innovation.

The Future Embryologist

Their role is changing and there are aspects of an embryologist’s work that are more obviously suited to being delegated to AI, such as measuring follicles or counting the number of cells in an embryo. However, healthcare professionals aren’t data scientists and they need to understand AI before they’d be willing to introduce it into a clinical setting. A lot of the embryologists we spoke to feel a sense of mystery and fear of AI. 
With the right education and time, AI can be pitched as another tool to help increase efficiency and improve IVF outcomes. AI is here to only enhance their practices and not replace their role, it has the potential to free them up to perform other embryology tasks that can only be performed by humans.

Show Your Workings

A common complaint of AI is it tends not to show its workings. So it will arrive at a decision or output, but it isn’t possible to review how it came to that decision. Known as the ‘black box’ effect, this can be down to the complexity of the algorithm and because the algorithms are proprietary so workings cannot be publicly shared. These are all hurdles to the clinical usage of AI and can be overcome by choosing more transparent and interpretable models. For AI to become a trustworthy tool to the professionals who are considering using it, this will be key. 


For clinicians and patients to trust the outputs of AI models there must be rigorous transparency and reporting of results. The effects of AI in medicine will affect all doctors and patients, not just those in the IVF sector, and the regulatory framework will continue to develop. We must therefore encourage open discussion on both the potential risks and benefits of this new technology. 


We need more data points for the mainstream use of AI in clinics, this will happen with time. With more than 3 million women globally having IVF each year, the more data that can be shared the the more effective AI models can be, but this needs to be done in a fair and medically confidential way. New methods to streamline data processing and digitise the vast reams of data collected during a single IVF cycle are also needed.  

Takeaways from the use of AI in IVF

AI in the IVF lab is just the beginning of how AI could improve the IVF experience for both patients and practitioners. From automating workflows to timely interactions with patients, this will save time and ultimately cost. Combine that with the growing concern of a shortage of IVF physicians, and the increasing global demand for IVF, automation of tasks becomes more important.

There is no doubt that IVF is a complex process involving several sequential decisions by physicians and embryologists. We need to maximise the potential for the success of all of those decisions and we believe AI can complement current clinical practices: 

  • AI is here to stay and clinics are starting to consider how to incorporate it into their practice. How we responsibly do this is a key focus. It’s not a case of whether AI will become part of your IVF cycle, but rather when. 
  • The opportunity for AI is considerable in an area that needs more objective and standardised decision-making. It will become an important tool in the IVF clinic’s toolbox to aid in limiting that human error and subjectivity. Regulators and clinics will have to work together to garner the large IVF data sets required for this to be possible.
  • Some studies suggest that AI has the potential to improve the accuracy and efficiency of IVF treatment, leading to higher success rates and fewer risks for patients. More research is needed to ensure that it is used ethically and responsibly.
  • Done right, AI should make IVF cheaper and therefore a more accessible treatment for those requiring medical intervention to bring a baby home. 
  • Not all IVF clinics in the UK are using AI technology yet, and those that do may use it in different ways or for different purposes. Looking for a clinic that does incorporate A.I into their IVF treatment? We’re compiling a list so please let us know here if your clinic uses AI.
  • ​​The HFEA would eventually like to offer a ‘regulatory sandbox’: a kind of controlled experiment in which fresh ways of doing things are trialled in a small-scale real-world environment. This could be ideal for AI and introducing it safely into the IVF clinic and you can read more about that in our Your Fertility Law Needs You! blog.

Recommended articles

  1. 09 July 2024

    OVUM's Guide to Fibroids and Fertility

    It’s Fibroids Awareness Month! So let's dive into the world of fibroids and e...

  2. 25 June 2024

    OVUM’s Guide to Cervical Cancer Screening

    If you're here, you're potentially thinking about cervical cancer screening (...

  3. 19 June 2024

    PCOS & Endometriosis: What You Need to Know

    One in 10 women have endometriosis; which, in the U.K alone, equates to aroun...

  4. 28 May 2024

    What It Means To Be PIF TICK Verified

    Think of the Patient Information Forum (PIF) TICK as the gold standard in hea...