Page 12 - AIH-2-3

Artificial Intelligence in Health AI in embryo selection for ART

identical studies to reduce bias. In addition, an experienced 3.6. Implications of the area under the curve (AUC)
professional reviewed the work to identify and address any in embryo selection
potential inconsistencies or biases. In embryo selection algorithms, the term “area under
Figure 3A illustrates the journal ranking of the the curve” refers to the area under the receiver operating
chosen studies, showing a clear dominance of Q1 (16) characteristic curve, which helps in assessing the probability
and Q2 (10) journals, with a smaller representation of successful implantation. AUC is a useful statistic for
24
from Q3 (1) and other categories (3). This distribution evaluating the performance of embryo selection algorithms,
highlights a significant representation of high-quality but it has several limitations, such as its reliance on image
articles in the selected literature. The dominance of Q1 quality, issues with generalizability, the impact of cultural
and Q2 journals reflects the intent to prioritize sources conditions, sex-dependent performance, and limits related
known for rigorous peer-review procedures and credible to sample size and research design. Even though AUC is
academic contributions, thereby assuring the reliability frequently used to assess model performance in embryo
of the findings. Meanwhile, Figure 3B presents the selection, depending solely on it presents challenges due to
publication trends over the past decade. Notably, 25 of sampling procedures and information transfer issues, which
the selected papers were published in the past 3 years, may affect the robustness and generalizability of the model.
indicating a recent increase in research efforts. In both AUC provides a general measure of model performance
2022 and 2023, eight publications were published, across all thresholds but does not account for individual
exhibiting continuous productivity. Surprisingly, the clinical contexts or specific requirements. Furthermore, it
highest point occurred in 2021 with nine publications, assumes that all misclassifications are equally important,
indicating a highly fruitful year for research in this whereas, in embryo selection, some misclassifications
domain. (e.g., false negatives) may have more severe consequences.
In addition, AUC does not evaluate the calibration of
3.5. Summary of the characteristics of the included predicted probabilities, which is a crucial requirement for
studies decision-making. Thus, AUC should be complemented
The current review thoroughly examined 30 academic with additional measures and clinical judgment for a more
publications that strictly adhere to formal academic comprehensive embryo assessment. 24,35-37
guidelines. Each paper features a clear and distinct title
and is authored by well-known researchers in the field. 3.7. Conventional study
These articles included abstracts that briefly describe their Any process or therapy that involves manipulating oocytes
goals, methodologies, and conclusions, as summarized (immature ova or egg cells) in vitro is referred to as ART
in Table 3, and have been published in reputable journals for reproduction. Couples and individuals experiencing
38
and conferences. The research methods spanned surveys, fertility issues can benefit from this treatment option,
case studies, and experimental studies. In addition to which is characterized by individualized treatment
reporting empirical results, the current review also protocols and multidisciplinary team management, both
offers critical comments on methodological challenges, of which improve treatment outcomes and safety. The
39
interpretative insights, and directions for further field of traditional ART has significantly advanced over
research. time. Techniques used in ART treatments include embryo

A B

Figure 3. Quality appraisal of the included studies. (A) Ranking statistics of the journals selected for this study. (B) Number of published papers in the
last 10 years.

Volume 2 Issue 3 (2025) 6 https://doi.org/10.36922/aih.7170