We meet three of the pioneers behind this cutting-edge ai@cam collaboration – a physicist, a fertility doctor and an infant neuroscience researcher – who are using AI to boost IVF success rates, tailor fertility treatments and deliver instant foetal health updates to expectant parents.

• Dr Staci Weiss, Department of Psychology, University of Cambridge
• Mo Vali, Department of Physics, University of Cambridge
• Dr Saaliha Vali, The Lister Fertility Clinic, Imperial College London

What were the challenges that inspired From womb to world: Revolutionising women’s health, fertility and early infant neurodevelopment using AI?

We are responding to a birth rates crisis, exacerbated by poor birth outcomes for couples who are finding it difficult to have children. In many advanced economies, the global fertility rate has more than halved over the past 50 years from about five children per woman to 2.1 today.

The key metric here is the replacement rate – the average number of children each woman would need to have to maintain a stable population size without immigration. Today, that number is lower than 2.1 in more than 110 countries. This trend is particularly worrying in places like South Korea, where the rate is 0.72 children per female (0.55 in Seoul), meaning the population is expected to fall by 60% in 70 years. In the UK, the fertility rate was 1.49 last year, and is predicted to fall to 1.38 in the next 20 years.

Rates are falling due to various factors: delay in younger, poorer women having children (with fewer children overall); increased female workforce participation; access to contraception; cost of living and childcare costs; male infertility; obesity and many other multi-factorial trends.

With fewer babies born and people living longer, many western economies face a rapidly ageing and shrinking population, leading to a real demographic crunch – with fewer young people generating wealth to support an elderly population. Plummeting fertility and live birth rates will change the global economy in ways we probably don’t even realise yet. And the problem is becoming worse. According to a paper published in the The Lancet in March 2024, 97% of the planet – 198 out of 204 countries – will have fertility rates below what is necessary to sustain their population by the end of this century.

Some governments are trying to support parents, offering financial incentives and other benefits. In South Korea, for example, there are companies offering up to $70,000 per child each time a woman has a baby, with no strings attached. Large tech companies such as Google are offering the chance for women to freeze their eggs. Some people simply don’t want children, while others face difficulties conceiving due to delayed childbearing, often resorting to in vitro fertilisation (IVF). Since 1978, about 12 million babies have been born through IVF, and an IVF baby is born roughly every 45 seconds.

IVF remains largely a numbers game

For those people who want to start a family, the current technology is inadequate. IVF outcomes are poor. One in four IVF cycles result in a live birth, and many women try IVF multiple times. Each cycle may involve one or more embryo transfers, taking months, plus recovery time. This makes IVF a very clunky process. IVF involves numerous internal ultrasounds, dozens of hormone injections, and other physically demanding procedures. It also costs upwards of £10,000 per cycle ($20,000 in America), so it’s a very big gamble to take.

Unfortunately, couples eager to start a family can fall prey to misinformation, willing to pay whatever it takes for the slight probability that a procedure or test can help them conceive. It doesn’t help that, scientifically, this area is not well understood. Women’s health has been neglected for decades; in the US, women only started being included in clinical trials in the 1990s. In the UK, only about 2% of medical research funding goes towards pregnancy, childbirth, fertility and women’s health. In fact, the UK has one of the largest gender health gaps among the G20 countries, which prompted the government to launch a Women’s Health Strategy for England in 2022.

How will the ai@cam project work to address this birth rate crisis?

The ai@cam project is seeking to develop cheaper, less invasive and more accurate AI-assisted tests that can be used throughout the conception to childhood journey. We want to improve the accuracy of early diagnosis of women’s health conditions, personalise fertility and IVF outcomes – and support the transition to parenting.

Our vision is to develop a new, holistic monitoring and treatment pathway – from the preconception to embryonic, neonatal and infant stages. Using AI, we want to help shed light on the still largely unknown underlying cellular, molecular and genetic mechanisms for how a life takes shape, including the genesis of neurodevelopmental disorders like autism. These tools will align with the needs of clinicians on the frontline, helping to diagnose conditions for women and couples alike.

Starting with fertility, we hope to transform the results of assisted reproduction outcomes by helping around the 55,000 women in the UK who are currently unable to have children naturally.

Using AI, we can personalise fertility treatments based on each woman’s profile (including race, age and medical history), bringing a new level of personalisation to healthcare. We’ll be using cutting-edge imaging and computer vision algorithms to help detect anatomical features and biomarkers in the foetus, allowing for early diagnosis of diseases and, ultimately, reducing the costs for the National Health Service (NHS).

What will the research look like?

We’re starting by collecting vast amounts of data from fertility clinics like ultrasound images, blood samples and follicular fluid, as well as demographic data such as the patient’s age, body mass index (BMI) etc. The goal is to combine all this information to create personalised treatment plans for women undergoing IVF.

Right now, treatments are highly operator-dependent, and there’s too much subjectivity in how treatments are decided. We want AI to bring a consistent, objective approach, moving beyond the currently low success rates for IVF by making the process more efficient, more accurate and less invasive.

Our project will look at ultrasound images over time to identify features linked to successful pregnancies. With higher accuracy, we hope to offer women a much clearer idea of their chances, helping them make informed decisions.

We’re also developing tools to give parents real-time insights into their baby’s development, once they’re pregnant. Using cutting-edge 4D ultrasounds, we can track foetal movements and provide visual updates to parents. This continuous monitoring helps to bridge the gap between the 20-week NHS scan and birth. It’s about giving parents peace of mind and a deeper connection to their baby’s growth. Using computer vision, we create annotated videos, showing how the baby’s movements in the womb compare to their postnatal movements. The goal will be to capture early behaviour we see in the foetus, like thumb-sucking and yawning, again after birth in the newborn.

A recent article connecting foetal activity to infant neurodevelopment was published by a team including Staci in Nature Human Behaviour.

Ultimately, we’d like to create an online dashboard to give people a precise timeline of their pregnancy and baby’s development. We believe a personalised interface that helps parents-to-be track their health information will enrich families’ understanding of their own data. By involving expectant parents at every stage of our digital health tool’s development, we hope the ultimate tool will serve as a powerful way to support families through the entire journey, from conception to infancy. You can keep track of our progress at www.wombs2world.com.

Are there any ethical risks when it comes to using AI for maternal and childhood health?

Yes, definitely. The key concerns are bias, regulation and data security. The effectiveness of our algorithms is limited by the nature and diversity of the population we’re collecting samples from, which introduces a layer of sample bias. We’re focusing our data collection on one IVF clinic in one city (London), in one country, ignoring the heterogeneity we might encounter elsewhere. In the future, we need to partner with multiple clinics across different geographies to achieve a more diverse cohort.

In terms of regulation, there is currently no framework for necessary standards on how to properly deploy AI-based systems in reproductive health, which needs urgent attention from regulatory bodies.

The third concern is data privacy. We handle highly confidential data and we’ve addressed this concern by receiving funding to buy our own GPU (Graphics Processing Unit). This ensures that all our data processing is done on a separate server, rather than on third-party platforms. We’ve ensured all data is anonymised at source upon collection, with highly restricted and monitored access. There is no way to identify patients from our data – ages are in ranges, BMIs are in ranges, and there are no names. This helps to head off any data privacy concerns.

What inspired your own work in this field?

Mo Vali: “I’m a physicist, interested in exploring fundamental questions. Most IVF innovations have been commercially led, tinkering around the edges and trying to make technical improvements to the IVF process. However, there’s a lot of unexplored territory, with open questions about the underlying molecular mechanisms of life. We still don’t really know the underlying mechanisms for how life takes shape, and we hope to put the science on a more rigorous footing – improving outcomes, reducing misinformation and exploring some of these fundamental questions.”

Saaliha Vali: “As an Obstetrics and Gynaecology doctor, I’ve spent considerable time with women experiencing unexplained infertility. I was shocked at how inefficient the IVF process is and felt that, with all the technology around us, we should be using cutting-edge developments in healthcare. I was always quite excited about AI, but as a clinician, I lacked the skill set to take it any further. Talking to my brother Mo, who is doing a PhD in physics and machine learning, I started discussing these ideas, leading to this cross-fertilisation between myself as a fertility doctor and scientists like Staci and Mo.”

Staci Weiss: “I’m an IVF baby myself from an early generation of the technology. My parents were screened for a genetic disorder, and subsequently enrolled in clinical trials that led to my conception through assisted reproduction. So that was one example of making fertility accessible through research. I became a sonographer to fund my undergraduate degree. While I was studying particles, I realised that a lot of principles from physics apply to people too. I went on study child psychology and neuroscience, which sheds light on what babies can do and understand before they can speak. There’s data we can extract that abides by prediction principles that organise us from cells to organs to bodies, which can help us to connect conception to childhood. Now in my new role at University of Roehampton, I’ll learn from babies – we will follow up babies conceived via IVF through online surveys and computer vision analysis of parent-provided ultrasound and infant video recordings.”

Why does research like this matter?

Mo, Staci and Saaliha: Through the ai@cam collaboration, we aspire to make Cambridge a pioneer in AI-driven, cutting-edge research across fertility, reproductive health and foetal neurodevelopment, exploring fundamental scientific questions in these areas.

We want to build on Cambridge’s long tradition of research in this field. Indeed, the world’s first IVF clinic, Bourn Hall Fertility Clinic, was established in Cambridge by IVF pioneers Robert Edwards (a Professor at Churchill College), Patrick Steptoe (a medical doctor) and Jean Purdy (a clinical embryologist) – leading to the birth of the world’s first baby conceived through IVF treatment in 1978.

We have already established world-class collaborations – with Addenbrooke’s Hospital and The Lister Hospital – which will allow us to access data, run pilot trials and ultimately look towards product commercialisation. We want to translate our research from lab to clinic and from womb to world by developing rigorous tests to improve clinical outcomes that will benefit the wider public – helping women make informed decisions on the individual success rates of their IVF treatments and providing options for less invasive tests.

As told to Vicky Anning

More about From womb to world: Revolutionising women’s health, female fertility and early infant neurodevelopment using AI

The project is a collaboration between the University of Cambridge’s Computer Science and Physics, among others – in partnership with two leading specialty hospitals – Addenbrooke’s Hospital and The Lister Hospital.

This initiative is actively seeking collaborations with IVF clinics and hospitals looking to address foundational questions in women’s, reproductive and infant health using state-of-the-art AI methods.

If you would like to find out more about this research project and discuss areas of potential collaboration, please email Mo at mv487@cam.ac.uk, Staci at smw95@cam.ac.uk and Sal at s.vali@nhs.net.