Apricity present fertility research at ESHRE 2022

Peruse our poster summaries below to explore our contributions to this year's conference.

Poster 1: Factors associated with live birth rate (LBR) and multiple live birth rate (mLBR): HFEA vs ABM registries (led by Neringa Karpaviciute)

The HFEA (the Human Fertilisation and Embryology Authority) is the UK’s fertility regulator, overseeing treatment and reporting statistics on rates of success. In France, the ABM (the French Biomedicine Agency) is the equivalent. 

When it comes to fertility treatment, different countries have different policies in place and often a lack of legislation to dictate which protocols should be used. An example is the number of embryos transferred within treatment such as IVF - a factor known to affect the risk of multiple birth rate (when multiple babies, like twins, are born). 

In this research, Neringa Karpaviciute looked at the UK and French registries to identify the factors which affect birth rates and multiple birth rates. The study found that while the registries had similar views on the effects of age, there was disagreement with other major factors. There are also discrepancies on the reporting of data, which makes it very difficult to compare and learn from data on a large scale. 

We hope that studies like this will encourage standardisation of results and policies, so that we can learn from data as an industry and identify the higher success, lowest risk treatment for each patient. 

Poster 2: A multi-centre evaluation of a novel 4-cell embryo classification system based on intercellular contact points (led by Rishabh Hariharan)

At Fertility 2022, Apricity data scientist Rishabh Hariharan presented work exploring the shapes of embryos and specifically, their cellular contact points - how and where cells touch each other within embryos. The work proposed a new classification for embryos to better understand which type of embryo contributes to IVF success.

The poster builds on this work, and found that the new classification system was well received in a clinical setting. 63% of embryos were unanimously recognised using the new system, which certain types of embryos significantly associated with greater success.

Rishabh’s work suggests that changing current embryo classification could improve the process of selecting embryos for use within IVF treatment by choosing the embryo shapes best associated with success.

Poster 3: ICM (inner cell mall) segmentation is impacted by several factors for humans as well for AI models but AI models show consistency (led by Celine Jaques)

Traditional embryo evaluation is often done by the eyes of qualified embryologists, making it a subjective and time-consuming process strongly affected by individual embryologist knowledge and experience. As a result, it impacts the quality of care provided to patients.

Grading embryos can be very useful in embryo quality assessment. To improve the objectivity of grading, this can be done by AI. 

For this piece of work, data scientist Celine Jacques looked at four structural embryo components which affect the grading of embryos. Difficulties identifying these four components leads to different evaluation of the embryo. 

Celine’s work suggests that these four components are significant and should be incorporated into embryologist training and AI embryo evaluation models. Overall, the use of AI in embryo evaluation would be a great tool to improve consistency, objectivity within treatment.

Poster 4: Predicting the Number of  eggs Retrieved from Controlled Ovarian Hyperstimulation with Machine Learning (led by Timothy Ferrand)

One of the first steps within the IVF process is ovarian stimulation, which involves using medication to stimulate the ovaries into producing more eggs. Not all fertilised eggs turn into healthy embryos, so this step increases the chances of success.

Controlled ovarian hyperstimulation (COH) is a way to medically induce the release of eggs. However, different patients often react to different COH protocols in different ways, in turn affecting IVF outcomes. Clinicians need to decide on suitable and cost-effective ovarian stimulation protocols for patients that maximise the chances of success whilst minimising the risk of complications. 

A number of previous studies have identified AI models that recognise the factors which influence COH success (such as age and hormone levels), but these studies have been limited, considering only a small number of variables in isolation. 

In this study, Apricity trained three AI models to predict the number of eggs retrieved depending on COH protocol. The models identified the most and least important factors within protocols, which can be found in the poster.

Predictive models like these can help clinicians optimise protocol based on individual patients, to ensure that everyone receives the best quality care. The study also provided evidence that it is possible for external researchers to provide clinically-relevant insights based on sensitive fertility data in a fully secure, transparent and trustworthy manner.

Apricity's senior data scientist Peter He presented their research on the 3D reconstruction of human preimplantation embryo development at#eshre2022.

This work serves as a first step toward unlocking data captured in IVF clinics for research into cell arrangement in preimplantation embryos. Moreover, the work may enable a more accessible assessment of cell arrangement thus revolutionising embryo selection. This in turn would make a really big difference in patient care and improve success rates.

Peter was recently shortlisted by the AI summit as AI Innovator of the Year and spoke at the CogX Festival and Leadership Summit. We can’t wait to see his talk.

We are so thankful to be in this unique position at the forefront of meaningful fertility research. By continuing our work, we hope to continue making fertility treatment more successful, transparent and convenient for everyone. 

To find out more about our research, publications and resources, visit the AI hub on our website.

Enjoyed this read?

Submit your email to join our newsletter

By signing up for our newsletter, you consent to receiving our marketing emails. You can unsubscribe at any time by clicking the link in the email. Please see our Privacy Policy for more information.

Linkedin logo