Page 23 - AIH-2-3

Artificial Intelligence in Health AI in embryo selection for ART

Figure 5. Future scope of assisted reproductive technology using artificial intelligence

the field of reproductive health is advancing significantly, opportunities. AI technologies, such as automatic embryo
as indicated by the proliferation of AI-related conference scoring algorithms (e.g., KIDScore D5 v3), Bayesian
abstracts and the commercialization of numerous AI networks, and shallow artificial neural networks (e.g.,
models. As an illustration, in countries with embryo multilayer perceptron and recurrent neural network),
3
selection restrictions, Levenberg–Marquardt neural have been applied in ART for embryo selection. These AI
networks that have been trained to utilize local binary models have demonstrated various levels of accuracy, with
patterns have exhibited promising outcomes in predicting some achieving up to 90% prediction accuracy for live
68
the quality of oocytes and embryos. This could improve birth probabilities. However, the reviewed studies are often
the technology used in assisted reproduction. It is crucial retrospective and conducted in single centers, limiting their
to address AI’s shortcomings, such as algorithmic bias and generalizability. In addition, the lack of comprehensive
the need for future research and clinical trials, to ensure its datasets and previous pregnancy information poses
successful integration into ART. Employing large training challenges to model performance. Throughout this paper,
90
datasets and robust models is recommended to surmount we demonstrate that AI has the potential to enhance
these challenges. embryo viability, improve live birth prediction accuracy,
personalize treatment plans, minimize human errors, and
5. Conclusion standardize IVF practices. To make advancements in the
Despite the extensive literature on AI applications in field of ART, future research must focus on creating AI
embryo selection, our study specifically focuses on models that are transparent, interpretable, and thoroughly
addressing existing gaps in this area. The endeavor to verified through randomized controlled trials. To overcome
revolutionize ART is at a critical turning point where current limitations, the generalizability of AI models can
the convergence of AI and precision in embryo selection be improved by diversifying training datasets and tailoring
intersects. This study explores AI applications in embryo models to different clinical settings. Integrating AI-driven
selection, highlighting advancements, challenges, and predictive analytics and real-time decision support systems

Volume 2 Issue 3 (2025) 17 https://doi.org/10.36922/aih.7170