The Building and Application of a Machine Learning Model on Predicting Euploid Blastocysts in IVF Treatments Without PGT-A.

Embryologists could not know whether blastocysts were euploid or not in in-vitro fertilization treatments without preimplantation genetic testing for aneuploidy. However, preimplantation genetic testing for aneuploidy technology which might impair the developmental potency of blastocysts was in strict limitation in China. Therefore, a non-invasive method was necessary for embryologists to select euploid embryos as far as possible. A machine learning model based on preimplantation genetic testing for aneuploidy results, development dynamics, morphology of blastocysts and female age was built by two-logistic regression analysis. The area under the receiver operating characteristic curve of the model was 0.877. The accuracy rate, precision rate, recall rate, f-score, true negative rate, false positive rate, false negative rate of the model when performed in independent data set were 77.35%, 77.49%, 72.55%, 74.94%, 81.55%, 18.45%, 27.45%, respectively. Besides, the false positive rate of the model on predicting euploid blastocysts was significantly lower than Gardner grade and KIDScore D5. When selected blastocysts by the model, the frozen embryo transfer outcomes improved significantly than the frozen embryo transfer outcomes which selected blastocysts by Gardner selection. The predictive model had potency in selecting euploid blastocysts in in-vitro fertilization treatments without preimplantation genetic testing for aneuploidy technology.