The use of artificial intelligence in embryo evaluation for IVF

By Lucy Hamer

Infertility is estimated to affect as many as 186 million people across the world (Szamatowicz, 2016), with approximately 9% of reproductive-aged couples struggling to achieve successful pregnancy following 12 months of unprotected intercourse (Inhorn & Patrizio, 2015). Couples are increasingly turning to assisted reproductive technologies (ART) such as in vitro fertilisation (IVF) for help (Farquhar et al., 2018), in which the mother’s oocyte is fertilised outside of the body and the resulting embryo transferred to the uterus following a brief period of observation (HFEA, 2021). Despite many technological advancements (Eskew & Jungheim, 2017) since the first IVF birth in 1978 (Begum, 2008), these expensive and lengthy treatments don’t come with a guarantee of successful pregnancy or live birth (Szamatowicz, 2016).

The main factor limiting the efficiency of IVF cycles is the success of implantation (Begum, 2008), with the majority of transferred embryos being rejected by the mother’s uterus as a result of failed attachment to the endometrial lining (Niederberger et al., 2018).  Implantation success rates are strongly influenced by embryo quality (Wang et al., 2019) making accurate embryo assessment vital for maximising pregnancy rates (Niederberger et al., 2018; Wang et al., 2019; Zaininovic et al., 2019). However, current embryo selection methods are not wholly reliable or consistent (Niederberger et al., 2018; Wang et al., 2019).

The highest quality embryo for transfer is typically selected based on morphological observation during the blastocyst stage (Wang et al., 2019; Zaininovic et al., 2019; Khosravi et al., 2019), meaning that the accuracy of embryo evaluation, and the likelihood of successful pregnancy, is highly dependent on the experience and knowledge of the embryologist involved (Wang et al., 2019; Zaininovic et al., 2019). These analyses are incredibly subjective (Wang et al., 2019; Zaininovic et al., 2019; Khosravi et al., 2019) and vary significantly between embryologists due to differing interpretations of what a high-quality embryo looks like (Khosravi et al., 2019; Gupta, 2021). This observer variation is well-documented, with a set of 5 embryologists only assigning the same classification to 89 embryos out of almost 400 in a recent study (Khosravi et al., 2020).

Artificial intelligence (AI) is a concept originating from the 1950s (Wang et al., 2019; Kaul et al., 2020) which has now invaded nearly every aspect of daily life (Wang et al., 2019). AI can be defined as the ability of machines to display forms of intelligence and execute tasks that would usually be expected of humans. There are masses of applications but some of the most prevalent examples include face recognition, product recommendations and, more recently, self-driving vehicles (Wang et al., 2019; Zaininovic et al., 2019). A particularly innovative subfield of AI is machine-learning (ML), in which sophisticated systems are able to detect patterns in vast amounts of data and generate algorithms that can used to make predictions when the system is presented with a similar, but unfamiliar, scenario (Wang et al., 2019; Gupta, 2021; Kaul et al., 2020). 

Recent developments in AI have created opportunities for its application to medicine (Wang et al., 2019), with these technologies proving particularly useful for medical imaging purposes (Zaininovic et al., 2019). Some current uses include detecting skin lesions (Khosravi et al., 2020), analysing cardiovascular magnetic resonance images (Kaul et al., 2020) and diagnosing cases of diabetic retinopathy (Khosravi et al., 2020). The insights provided by these automated systems can be used to support clinical decision making, increasing the rate of accurate diagnoses and enhancing patient care (Wang et al., 2019; Kaul et al., 2020; Jiang et al., 2017). Current evidence indicates that AI outperforms doctors when it comes to medical image analysis, and it is also estimated that incorporation of AI into clinical practise would substantially cut healthcare costs (Bohr & Memarzadeh, 2020).

With embryo image evaluation being pertinent to the success of an IVF cycle, AI has the potential to be incredibly advantageous in the fertility sector. An automated image identification system could take the subjectivity out of blastocyst analysis and provide a more quantitative, consistent assessment of embryo quality (Wang et al., 2019) whilst also reducing costs for the fertility centre and, subsequently, the patients (Gupta, 2021). Recently, a US-based research team described the application of a ML technique known as deep learning, which involves the construction of artificial neural networks that mimic the inner workings of the brain, to blastocyst image analysis. After being trained on over 10,000 images, the algorithm was able to classify embryos as good or poor-quality with an accuracy of 97%. The results were confirmed with validation datasets from additional clinics and the team concluded that the deep neural network, known as STORK, was able to select the highest quality embryos for transfer with an accuracy superior to that of experienced embryologists (Khosravi et al., 2020).

The use of AI in ART is not limited to the evaluation and selection of embryos. The same technologies could similarly be applied to assess sperm quality (Zaininovic et al., 2019; Gupta, 2021), select the most appropriate oocyte or predict the likely outcome of IVF (Wang et al., 2019). The incorporation of AI into the healthcare system promises to improve patient outcomes and also opens doors for personalised infertility treatment (Wang et al., 2019; Zaininovic et al., 2019). However, the technology is not without its shortcomings. There are worries that AI may one day replace healthcare workers (Wang et al., 2019; Jiang et al., 2017; Haidar, 2019), although this is unlikely to be a reality and instead it is generally believed that AI is most useful as a supplementary tool to support physicians in providing optimal patient care (Wang et al., 2019; Jiang et al., 2017; Haidar, 2019; Bohr & Memarzadeh, 2020). 

There are also concerns surrounding bias. AI bias exists in aspects of daily life, with women being presented with fewer high-paying job vacancies compared to males (Curchoe & Bormann, 2019) whilst African Americans are more likely to be denied loans than Caucasians (Panch et al., 2019). Bias is particularly likely when the datasets used to train the algorithms are insufficient or biased themselves (Wang et al., 2019), so care needs to be taken to ensure that training datasets are faultless.

AI undoubtedly has the potential to revolutionise every aspect of healthcare and is particularly applicable to embryo evaluation and selection within the process of IVF. Despite being in existence for almost 75 years, these technologies are still fairly experimental, and more research is required to ensure that the integration of automated algorithms into diagnosis and treatment workflow pathways, will not result in adverse consequences for the patients, physicians or healthcare sector as a whole.


Szamatowicz M. Assisted reproductive technology in reproductive

medicine — possibilities and limitations. Ginekologia Polska. 2016: 87(12); 820–823. Available from doi: 10.5603/GP.2016.0095

Inhorn MC & Patrizio P. Infertility around the globe: new thinking on gender, reproductive technologies and global movements in the 21st century. Human Reproduction Update. 2015: 21(4); 411-426. Available from doi: 10.1093/humupd/dmv016

Farquhar C, Marjoribanks J, Cochrane Gynaecology and Fertility Group. Assisted reproductive technology: an overview of Cochrane Reviews. Cochrane Database Syst Rev. 2018: 2018 (8); CD010537. Available from doi: 10.1002/14651858.CD010537.pub5

Human Fertilisation & Embryology Authority. In vitro fertilisation (IVF). Available from: [Accessed 1st March 2021]

Eskew AM & Jungheim ES. A History of Developments to Improve in vitro Fertilization. Missouri Medicine. 2017: 114(3); 156-159. Available from: [Accessed 1st March 2021]

Begum MR. Assisted Reproductive Technology: Techniques and Limitations. Journal of Bangladesh College of Physicians and Surgeons. 2008: 25; 135-141. Available from doi: 10.3329/jbcps.v26i3.4197

Niederberger C, Pellicer A, Cohen J, Steirteghem AV, Rebar RW, LaBarbera AR. Forty years of IVF. Fertility and Sterility. 2018: 110(2); 185-324. Available from doi: 10.1016/j.fertnstert.2018.06.005

Wang R, Pan W, Jin L, Li Y, Geng Y, Gao C, et al. Artificial intelligence in reproductive medicine. Reproduction. 2019: 158(4); 139-154. Available from doi: 10.1530/REP-18-0523

Zaininovic N, Elemento O, Rosenwaks Z. Artificial intelligence: its applications in reproductive medicine and the assisted reproductive technologies. Fertility and Sterility. 2019: 112(1); 28-30. Available from doi: 10.1016/j.fertnstert.2019.05.019 

Khosravi P, Kazemi E, Zhan Q, Malmsten JE, Toschi M, Zisimopoulos P, et al. Deep learning enables robust assessment and selection of human blastocysts after in vitro fertilization. Npj Digital Medicine. 2019: 2; 21. Available from doi: 10.1038/s41746-019-0096-y

Gupta S. AI will revolutionize assisted reproductive technology (if we work together). Available from: [Accessed 1st March 2021]

Kaul V, Enslin S, Gross SA. History of artificial intelligence in medicine. Gastriointestional Endoscopy. 2020: 92(4); 807-812. Available from doi: 10.1016/j.gie.2020.06.040 

Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, et al. Artificial intelligence in healthcare: past, present and future. Stroke and Vascular Neurology. 2017: 4(2). Available from doi 10.1136/svn-2017-000101

Bohr A & Memarzadeh K. The rise of artificial intelligence in healthcare applications. Artificial Intelligence in Healthcare. 2020: 25-60. Available from doi: 10.1016/B978-0-12-818438-7.00002-2 

Haidar H. Benefits and challenges of using artificial intelligence in assisted reproductive technologies. Available from: [Accessed 1st March 2021]

Curchoe CL & Bormann CL. Artificial intelligence and machine learning for human reproduction and embryology presented at ASRM and ESHRE 2018. Journal of Assisted Reproduction and Genetics. 2019: 36; 591-600. Available from doi: 10.1007/s10815-019-01408-x

Panch T, Mattie G, Atun R. Artificial intelligence and algorithmic bias: implications for health systems. Journal of Global Health. 2019: 9(2); 020318. Available from doi: 10.7189/jogh.09.020318

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s