Artificial Intelligence in the service of Intra Uterine Insemination and Timed Intercourse in Spontaneous Cycles

To develop a machine learning model designed to predict time of ovulation and the optimal fertilization window for performing intrauterine insemination or timed intercourse in natural cycles. A retrospective cohort study. A large IVF unit. Patients undergoing 2467 natural frozen embryo transfer cycles between 2018-2022. None MAIN OUTCOME MEASURE: Prediction accuracy of the optimal day for performing Insemination or timed intercourse. The dataset was split into a training set including 1864 cycles and two test sets. In the test sets ovulation was determined according to either expert opinion, with two independent fertility experts determining ovulation day ("expert") (496 cycles), or according to the disappearance of the leading follicle between two consecutive days ultrasound exam ("certain ovulation") (107 cycles). Two algorithms were trained: 1) An NGBoost machine learning model estimating the probability of ovulation occurring on each cycle day 2) A treatment management algorithm using the learning model to determine an optimal insemination day, or whether another blood test should be performed. Estradiol progesterone and LH levels on the last test performed were the most influential features used by the model. The average number of tests was 2.78 and 2.85 for the "certain ovulation" and "expert" test sets respectively. In the "expert" set the algorithm correctly predicted ovulation and suggested day-1 or -2 for performing insemination in 92.9% of cases. In 2.9% the algorithm predicted a "miss" meaning the last test day was already ovulation day or beyond, suggesting avoiding performing insemination. In 4.2% the algorithm predicted an "error", suggesting to perform insemination when in fact it would have been performed on a non-optimal day (0 or -3). The "certain ovulation" set had similar results. This is the first study implementing machine learning model, based on blood tests only, for scheduling insemination or timed intercourse with high accuracy, attributed to the capability of the algorithm to integrate multiple factors and not rely solely on the LH-surge. Introducing the models' capabilities may improve accuracy and efficiency of ovulation prediction and increase chance of conception.