Key points are not available for this paper at this time.
Abstract Study question Is AI-based evaluation of embryos as effective in achieving pregnancy as the standard assessment performed by an experienced embryologist? Summary answer AI-based assessment of blastocysts may result in comparable pregnancy rates as the standard assessment performed by an experienced embryologist. What is known already Standard clinical practice in the IVF procedure includes morphological assessment of blastocysts based on visual evaluation following Gardner scale. However, this subjective grading process is susceptible to variability dependent on the professionals’ experience and knowledge, leading to potential inconsistencies in selecting embryos for transfer. Recently, there has been a suggestion that artificial intelligence (AI) algorithms could enhance the predictability of treatment outcomes and many AI-based software applications have been introduced to the procedure in many IVF centers. Nevertheless, there are very few prospective evaluations of the real-life implementation of the novel technologies. Study design, size, duration It is an ongoing, randomized, prospective, multicenter, international study performed in four investigation centers since November 2023. The duration of the study is scheduled for one year. Participants/materials, setting, methods The first stage of the evaluation comprises a randomized selection of a patient meeting the inclusion criteria to one of the groups: test or control group. The next step includes a selection of the embryo for implantation. For the control group, the experienced embryologist performs the standard selection of embryos as per normal clinical practice following the Gardner scale. For the study group, the embryologist selects the embryo for transfer indicated by the EMBRYOAID application. Main results and the role of chance Currently, there have been 149 patients recruited in the study and 160 cycles with 638 embryos have been entered in the EMBRYOAID software for AI-based evaluation. Test group of 80 patients comprises 87 cycles with 346 embryos. Preliminary results of the prospective study conducted in real-life settings show encouraging outcomes in relation to the idea that AI-supported embryo evaluation may be as effective in regards to obtaining pregnancy as a standard clinical assessment performed by experienced professionals. Nevertheless, the sample size is not yet sufficient to achieve statistical significance. Therefore, further data gathering will be conducted and the study hypothesis will be furthermore examined. Limitations, reasons for caution The sample size is still insufficient to reach statistical significance and the study shows the preliminary results which will be further verified on a larger patient group. Wider implications of the findings Results of a prospective, randomized trial conducted in several international clinics may enhance the trend to introduce more objective, bias-free AI-based tools into the embryology labs. Trial registration number EMBRYO National Center for Research and Development nr POIR.01.01.01-00-1631/20-00; ERC Consolidator Grant TUgbOAT no 772346
Wygocki et al. (Mon,) studied this question.