The following is a summary of “Deep learning for embryo evaluation using time-lapse: a systematic review of diagnostic test accuracy,” published in the NOVEMBER 2023 issue of Obstetrics and Gynecology by Berman, et al.

For a study, researchers sought to assess the accuracy of convolutional neural network (CNN) models in evaluating embryos through time-lapse monitoring.

A systematic search was conducted in the PubMed and Web of Science databases from January 2016 to December 2022, utilizing keywords and MeSH (Medical Subject Headings) terms. Studies reporting the accuracy of CNN models for embryo evaluation using time-lapse monitoring were included. The review was registered with PROSPERO. Two reviewers independently screened results using Covidence systematic review software. Full-text articles were reviewed when studies met the inclusion criteria or in case of uncertainty. A third reviewer resolved Nonconsensus. The QUADAS-2 tool and the modified Joanna Briggs Institute (JBI) checklist were employed to evaluate the risk of bias and applicability.

Following a systematic literature search, 22 eligible studies were identified, all of which were retrospective. A total of 522,516 images of 222,998 embryos were analyzed. Three main outcomes were assessed: successful in vitro fertilization, blastocyst stage classification, and blastocyst quality. Most studies reported accuracy exceeding 80%, with some outperforming embryologists. Ten studies had a high risk of bias, primarily due to patient bias.

Integrating artificial intelligence in time-lapse monitoring holds the potential for more efficient, accurate, and objective embryo evaluation. Models focusing on blastocyst stage classification demonstrated the most reliable predictions. Models predicting live births exhibited low bias risk, utilized extensive databases, and underwent external validation, enhancing their clinical relevance. The study’s limitations included high heterogeneity among the included studies, emphasizing the need for shared databases and standardized reporting by researchers.

Source: ajog.org/article/S0002-9378(23)00264-8/fulltext