Embryo selection through artificial intelligence versus embryologists: a systematic review

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Embryology

Background:

• In vitro fertilization (IVF) success rates remain suboptimal, particularly with increasing female age.
• Artificial intelligence (AI) offers potential to enhance objectivity in embryo selection, a process traditionally limited by human subjectivity.
• Reviewing AI's performance against embryologists in embryo selection is crucial for improving IVF outcomes.

Purpose of the Study:

• To compare the performance of artificial intelligence (AI) decision support systems with human embryologists in embryo selection.
• To evaluate AI's accuracy in predicting embryo morphology and clinical outcomes in IVF treatments.

Main Methods:

• A comprehensive literature search was conducted across major scientific databases (PubMed, EMBASE, Ovid Medline, IEEE Xplore) from 2005 to 2022.
• Included studies focused on AI applications in IVF, assessing accuracy, sensitivity, and specificity of embryo assessment and clinical outcome prediction.
• Risk of bias was evaluated using the Modified Down and Black Checklist.

Main Results:

• AI models demonstrated higher accuracy than embryologists in predicting embryo morphology (median 75.5% vs. 65.4%) and clinical pregnancy (median 77.8% vs. 64%).
• When combining image/time-lapse and clinical data, AI achieved a median accuracy of 81.5%, significantly outperforming embryologists (51%).
• AI consistently showed greater reliability and accuracy in predicting both embryo quality and IVF success.

Conclusions:

• AI shows significant promise for improving embryo selection in IVF, consistently outperforming clinical teams in assessed studies.
• Further research is needed to prospectively evaluate AI in clinical settings and address model generalizability through external validation.
• A shift in focus towards live birth as the primary clinical outcome and the use of diverse, validated databases are recommended for AI development in IVF.