Intelligent assisted reproduction: Innovative applications of artificial intelligence in embryo health assessment

Infertility is a widespread global concern, affecting approximately 17.5 % of adult couples worldwide. Assisted reproductive technologies (ARTs), particularly in vitro fertilization (IVF), have emerged as a promising solution for millions of couples facing infertility. However, the IVF success rate remains low, averaging around 30 %.The quality of the selected embryos is a key determinant of successful embryo transfer, indicating the importance of precise and accurate assessments of embryo viability in improving IVF outcomes.

Traditional methods for embryo health assessments primarily involve morphological evaluations under a microscope and the examination of various factors such as quality of the inner cell mass (ICM) and trophectoderm (TE). The accuracy (ACC) of these assessments depends largely on the expertise of embryologists and external conditions within the laboratory environment. Embryologists’ interpretations of morphological features can vary because of differences in visual perception and evaluation criteria. For instance, some embryologists may overlook minor irregularities, whereas others might assign lower grades for the same features. Moreover, variations in microscope quality and lighting conditions can lead to inconsistent interpretations of the same embryo by different embryologists.

Another limitation of traditional embryo health assessments is the lack of precise quantification. Although indicators such as cell symmetry and fragmentation rate are evaluated, these assessments are primarily visual and lack standardized, quantifiable metrics, introducing subjectivity into the results. Recent advancements of the artificial intelligence (AI) and imaging technologies offer promising solutions in this domain. AI-driven deep learning algorithms can process large datasets of embryonic images to generate accurate and objective evaluations. These approaches enhance consistency and enable continuous monitoring of embryo development through advanced techniques such as time-lapse (TL) imaging, resulting in more detailed assessments. By minimizing reliance on individual expertise and subjective interpretation, AI offers new opportunities to clinicians in making evidence-based decisions when selecting the most viable embryos.

As AI adoption in ART and clinical practice continues to evolve, this review systematically examines its application in assessing embryo health at different developmental stages. This review also evaluates the feasibility of AI-based approaches compared with traditional embryologist-led assessments. Furthermore, this review explores the applicability of AI in diverse clinical settings, its data quality requirements, ethical considerations, and potential directions for future development.

An independent literature search was conducted using academic databases such as Google Scholar and Semantic Scholar by employing keywords including “artificial intelligence,” “embryo health assessment,” “IVF,” and “embryo selection.” The article screening process is illustrated in Fig.1. This review exclusively focuses on studies examining embryo health assessment and selection, while excluding studies that specifically targets the prediction of live birth outcomes or embryo euploidy. A total of 37 studies were included: 11 addressed early developmental stages, 15 examined the blastocyst stage, and 11 covered the full developmental timeline. The majority of the reviewed studies have been published between 2013 and 2024.