A fully booked Campus meeting on embryo selection reviewed a catalogue of emerging concepts and developments in embryo selection - notably covering PGT-A, time-lapse and artificial intelligence - but behind the presentations lay an agreement that morphological assessment still sets the gold standard.
The theme of a fully booked Campus meeting hosted by the SIG Embryology in Krakow, Poland, in May was ‘novel approaches to embryo selection’, but there was an underlying agreement in most of the presentations that, despite the strengths of time-lapse, PGT-A, artificial intelligence systems and automation in the lab, morphology remains the gold standard and sets the benchmark against which all emerging methods must be measured. Nevertheless, time-lapse in its dynamic imaging of embryo development has provided new insights - in compaction, blastulation, the speed of cleavage, and even in evaluation from static images - in the morphological assessment of embryos. All such developments were on the agenda of this Campus meeting.
Nevertheless, the SIG Embryology’s co-ordinator, Gemma Arroyo from Barcelona, acknowledged that embryo grading systems ‘required an update’, with routine benefits ‘proven’ but still with limitations. The essential checkpoints of morphological grading, she reaffirmed, were fertilisation and first cleavage at day 1, second and third cleavages to four cells at day 2, cleavage to eight cells at day 3, compaction at day 4, and blastulation at day 5. However, Arroyo also agreed that time-lapse systems offer assessment advantages over conventional benchtop incubators, even though with respect to live birth she claimed no strong evidence of better embryonic selection by morphokinetics than by conventional morphology. However, she did highlight the evidence (‘low quality’) in favour of blastocyst transfer over cleavage stage in fresh cycles, even though there was as yet no such evidence in cumulative cycles.(2)
She further noted that that many of those who still perform embryo transfer and freezing on day 3 are not yet users of new and progressive technologies such as time-lapse and AI. Indeed, a later talk by the SIG’s senior deputy Mónica Marques agreed that time-lapse findings may actually enhance current static morphology assessment.
Present grading systems, said Arroyo, are still largely based on recommendations from the influential Istanbul workshop of 2011, with more recent evidence from time-lapse data currently described as ‘complementary’.(3) That workshop had graded morula as ‘good’ if the embryo in its fourth round of cleavage had evidence of compaction within the whole embryo volume. Presentations from Thomas Ebner, Cristina Magli and Mónica Marques presented new evidence about morulation and blastulation, notably that late onset of vacuolisation around the compaction stage is a negative predictor of blastocyst formation and outcome, and that blastocyst selection might also depend on the inner cell mass (larger, more compact) and trophectoderm (homogeneous cell size).
Former SIG Embryology co-ordinator Giovanni Coticchio outlined other developments from time-lapse morphokinetic checkpoints in fertilised oocytes. Indeed, he emphasised that adopting the Istanbul consensus for parameters such as the timing of fertilisation and the presence of multi-pronuclear zygotes may result in the wastage of zygotes which might have led to a healthy live birth.
SIG deputy co-ordinator Amy Barrie in her presentation on time-lapse concluded that algorithmic embryo selection seems most predictive when associated with morphological selection - and that commercially available algorithms (rather than locally developed) appear to offer better predictive ability. For example, a very recent retrospective study of an embryo assessment algorithm applied on Day 3 to around 5000 embryos found that classification results were indeed associated with implantation and live birth.(4) Notably, highest performance was achieved when combining the automatic scoring system with traditional morphological classification.
However, this study found no association between embryo classification and euploidy status in embryos analysed by PGT-A, a common but contentious finding. Indeed, Barrie cited three studies from 2023 alone, with mixed results, one even suggesting that an automated system may indicate ‘how many and which blastocysts to biopsy’. Nevertheless, aneuploidy detection remains an elusive challenge for morphokinetics.
The best detector of ploidy status remains PGT-A via trophectoderm biopsy, but, as Elpida Fragouli said, this remains an expensive procedure and may not be indicated routinely in all patients. She reviewed developments in non-invasive methods, agreeing that they remain at the stage of ‘hope’, but not yet ‘reality’. The two approaches so far documented are from measurement of cell-free DNA in used culture medium and in the collection and analysis of blastocoel fluid. The latter, she described as ‘minimally’ and not ‘non’-invasive, because it still requires the collection of fluid from within the blastocyst.
The advantages of non-invasive methods, said Fragouli, were not so much that biopsy might cause damage to the cells but more that the technique requires great skill and adds cost to the patient. However, although non-invasive methods can be applied in any laboratory without need for extra equipment, there remain problems in correlating measurements of DNA with ploidy status - as determined by analysis of the DNA from the used culture medium and/or blastocoel fluid with that from a biopsy - and the possibility of contamination. Clinical results, she noted, are encouraging, but RCTs are still needed before making firm conclusions.
What about artificial intelligence in embryo selection? It’s the object of innumerable studies right now in gamete selection and embryo scoring and, according to Ioannis Sfontouris, past co-ordinator of the SIG Embryology, AI now lies behind the move towards ‘data-driven’ embryo selection. Indeed, ESHRE’s own recommendations on the use of time-lapse systems (from 2020) expressed ‘little doubt that the future of AI and TLT will incorporate some degree of machine learning, to facilitate complex analysis of large data sets, which will likely reveal currently unidentified combinations of visual markers’.(5)
For Sfontouris the potential applications of AI in IVF lie in the annotation of embryo development, embryo grading and selection, sperm analysis and selection, and the prediction of IVF outcome - with a plethora of studies already in the literature in each of these approaches. But, while individual results offer encouragement, he echoed the caution of others that AI is not yet ‘a miracle worker’. ‘It’s one thing to succeed in the lab, quite another to do so in real life,’ he said. Even so, and without the benefits of unequivocal evidence so far in outcomes, AI like time-lapse has already shown benefits in embryo selection - in improving consistency in evaluation, in time-saving and better workflow, and in training and quality control. An improvement in pregnancy rate and time-to-pregnancy from more accurate embryo selection seems sure to follow - although right now the evidence of clinical benefit is still lacking for both time-lapse and AI. But, asked Sfontouris and as one study has suggested, are we now heading towards the ‘datification, digitalisation, algorithmisation of ART’. Certainly, from this Campus meeting, the future of embryo selection lies not only in the hands and eyes of the embryologist but in the images, information and prediction which huge datasets can provide.
1. Armstrong S, Bhide P, Jordan V, et al. Time-lapse systems for embryo incubation and assessment in assisted reproduction. Cochrane Database Syst Rev 2018; 5(5):CD011320. doi.org/10.1002/14651858.CD011320.pub3
2. Glujovsky D, Quinteiro Retamar AM, Alvarez Sedo CR, et al. Cleavage-stage versus blastocyst-stage embryo transfer in assisted reproductive technology. Cochrane Database Syst Rev 2022; 5(5):CD002118.
3. Alpha Scientists in Reproductive Medicine and ESHRE Special Interest Group of Embryology. The Istanbul consensus workshop on embryo assessment: proceedings of an expert meeting. Hum Reprod 2011; 26: 1270-1283.
4. Valera MA, Aparicio-Ruiz B, Pérez-Albalá S, et al. Clinical validation of an automatic classification algorithm applied on cleavage stage embryos: analysis for blastulation, euploidy, implantation, and live-birth potential. Hum Reprod 2023;
5. ESHRE Working group on Time-lapse technology. Good practice recommendations for the use of time-lapse technology. Hum Reprod Open 2020; hoaa008 https://doi.org/10.1093/hropen/hoaa008