Artificial intelligence-assisted selective modified natural frozen embryo transfer

Can an artificial intelligence algorithm assist in selectively triggering natural-cycle frozen embryo transfer (NC-FET) to avoid transfers on undesired days?

This was a retrospective cohort study including patients undergoing 4029 FET cycles (2018-2024). The proposed algorithm used a previously developed ovulation prediction model that can predict ovulation up to 6 days in advance using LH, oestradiol, progesterone and follicle size. The algorithm predicted the probability of natural ovulation occurring in the days following each clinic visit and suggested triggering only in cases where a transfer was likely to occur on days the clinic would be closed.

The results of the current analysis represent a selected scenario for a clinic that is closed on Sundays and has minimal staffing on Saturdays. Other clinic schedules can be configured. In around 40% of cycles, the algorithm estimated ovulation occurring on a day leading to a non-working day transfer so triggering was suggested. These cases resulted in a shift of ovulation by 2 days or fewer in 89% of cases, and 4 or more days in 2% of cases. In the remaining approximately 60% of cases, the algorithm did not intervene in natural ovulation, identifying ovulation correctly in 96% of cases.

The use of a selective modified NC-FET algorithm can minimize the number of transfers performed on non-working days, while keeping the majority of cycles completely natural, enabling flexibility when needed, and maintaining the advantages of natural cycles when possible. Implementation of this algorithm may provide a solution for clinics looking to increase their NC-FET practice without compromising convenience.