An artificial intelligence model based on the proteomic profile of euploid embryos and time-lapse images: a preliminary study

Abstract Research Question To develop an Artificial Intelligence (AI) model based on Artificial Neural Networks (ANNs) to predict the likelihood of achieving a live birth (LB) by using the proteomic profile of spent culture media and morphological data from time-lapse images. Design This retrospective cohort study included 212 patients who underwent single blastocyst transfer in IVI Valencia. Image analysis was performed in 186 time-lapse (TL) blastocyst pictures and the protein profile was analyzed in 81 spent embryo culture media from patients included in the Preimplantation Genetic Testing (PGT) program. The extracted information from both analysis was used as input data for the ANN. The Multilayer Perceptron and the Back-propagation learning method were used to train the ANN. Finally, the predictive power was measured using the area under the curve (AUC) of the Receiver Operating Characteristic curve. Results Three ANN architectures classified most of the embryos correctly as LB+ and LB-: 100.0% for ANN1 (morphological variables and 2 proteins), 85.7% for ANN2 (morphological variables and 7 proteins), and 83.3% for ANN3 (morphological variables and 25 proteins). The AI model proposed with information extracted from TL blastocyst image analysis and levels of Interleukin 6 (IL-6) and Matrix metalloproteinase 1 (MMP-1) was able to predict LB on testing data with an AUC of 1.0. Conclusions The model proposed is a promising tool to select the most successful embryo of a euploid cohort. The high accuracy of prediction demonstrated by this software may improve the efficacy of an assisted reproduction treatment reducing the number of transfers per patient.