RESUMO
OBJECTIVE: To decrease multiple pregnancy risk and sustain optimal pregnancy chance by choosing suitable number of embryos during transfer, this study aims to construct artificial intelligence models to predict the pregnancy outcome and multiple pregnancy risk after IVF-ET. MATERIALS AND METHODS: From Jan 2010 to Dec 2019, 1507 fresh embryo transfer cycles contained 20 features were obtained. After eliminating incomplete records, 949 treatment cycles were included in the pregnancy model dataset and 380 cycles in the twin pregnancy model dataset. Six machine learning algorithms were used for model building based on the dataset which 70% of the dataset were randomly selected for training and 30% for validation. Model performances were quantified with the area under the receiver operating characteristic curve (AUC), accuracy, specificity, and sensitivity. RESULTS: Models built with XGBoost performed best. The pregnancy prediction model produced accuracy of 0.716, sensitivity of 0.711, specificity of 0.719, and AUC of 0.787. The multiple pregnancy prediction model produced accuracy of 0.711, sensitivity of 0.649, specificity of 0.740, and AUC of 0.732. CONCLUSIONS: The AI models provide reliable outcome prediction and could be a promising method to decrease multiple pregnancy risk after IVF-ET.