As women approach the end of their childbearing years, the incidence of aneuploid embryos increases, which often results in serious clinical consequences such as infertility, miscarriage, and birth defects. Due to this, couples are increasingly seeking assisted reproductive technologies to conceive.

Artificial intelligence (AI) has become increasingly useful in the field of assisted reproduction, with several studies attempting to develop unbiased and automated embryo selection tools using deep learning. However, one challenge that remains is predicting the ploidy status of embryos as a means of prioritizing and selecting embryos with the highest implantation potential.

Embryo selection is a crucial part of in vitro fertilization (IVF) treatments, where embryologists assess the quality of embryos by determining their ploidy status (the number of sets of chromosomes they contain) and then identify the best ones for implantation.
In a recent study, researchers have attempted to develop a deep-learning model that could assist in this process and prioritize embryos with the highest implantation potential.

The STORK-A deep-learning approach was used to classify the ploidy status of embryos in three distinct classification tasks: aneuploids versus euploids; complex aneuploid versus euploids plus single aneuploid; and complex aneuploid versus euploids.
The study found that the use of static images of embryos to predict ploidy status did not markedly improve the performance of STORK-A when comparing machine-learning models and deep-learning models across the three classification tasks.
It also identified several limitations, including the potential bias introduced by embryos being selected by embryologists as candidates for preimplantation genetic testing for aneuploidy (PGT-A) based on their morphology. Another limitation was the use of images captured only by time-lapse microscopy, limiting generalizability.

Despite these limitations, the study shows that AI can approach the performance of expert annotators in providing useful information about embryo implantation potential.
Future studies may incorporate temporal and spatial information from videos of embryo development to improve ploidy predictive performance.