Page 20 - AIH-2-3

Artificial Intelligence in Health AI in embryo selection for ART

is transforming the landscape of ART. This exploratory 4.3.1. RQ1: What are the current state-of-the-art AI
29
review examines recent developments in embryo technologies used in embryo selection for ART?
selection and related AI-driven innovations within ART. In ART, recent developments in AI have significantly

4.2. Strength of the systematic review improved embryo selection procedures. These cutting-
edge AI tools use ML, deep learning, and computer
The strength of this systematic review lies in its vision to increase the precision, reliability, and efficiency
thorough analysis of the correlation between AI and of embryonic health evaluation. AI technologies are
ART, particularly those concerning embryo selection. transforming IVF laboratories by automating the
The review thoroughly explains the opportunities, assessment of embryo morphology and leveraging
challenges, and future directions in using AI to improve synthetic data. The primary AI tools currently employed in
ART outcomes by analyzing a wide range of literature embryo selection are listed below:
from the past 10 years. This paper is the first to highlight (i) Computer vision and deep learning: Computer
the need for external validation of prediction models, vision and deep learning are used by AI systems
an aspect that requires significant improvement in to automatically examine images of embryo
the reviewed studies. Most of the scientific papers morphology and extract important aspects that are
used retrospective and anonymized data, which may essential for determining the survival of the embryo.
introduce biases and limit the generalization of findings. The accuracy of embryo selection is improved, and
Therefore, this systematic review combines an in-depth the subjectivity associated with manual assessments
analysis of various documents from several countries, by embryologists is reduced. For example, deep
92
thus reducing bias and highlighting its potential to learning algorithms that have been trained on both
transform the field of ART. It also highlights ethical synthetic and actual embryo images have shown
considerations and emphasizes the significance of excellent accuracy in predicting the stages of embryo
responsible implementation in clinical practice. With cells, reaching up to 97% accuracy when synthetic
its in-depth analysis and interdisciplinary approach, data is included. 93
the review provides valuable insights for researchers, (ii) Time-lapse imaging: AI systems use time-lapse
clinicians, and policymakers, facilitating informed imaging to improve predictions from fertilization to
decision-making. In addition, four key parameters the blastocyst stage, thereby increasing IVF success
have been identified – ethical concerns, clinical and rates by identifying viable embryos more accurately
regulatory constraints, discussion of AI techniques, and than human experts. 94
the potential applications of AI in ART – to illustrate the (iii) ML Techniques: Several ML techniques, such as neural
strength of this review (Table 4). networks, naive Bayes, support vector machines,
and random forests, are used to predict treatment
4.3. Addressing the RQs outcomes and enhance IVF results. The average

In subsequent sections, comprehensive responses are AUC rating for these models is 0.91, indicating high
provided to address the RQs. accuracy, sensitivity, and precision. 94

Table 4. A thorough comparative analysis of relevant studies and the current review

References Ethical Regulatory and AI techniques Prospects of
concerns clinical barriers discussed AI in ART
Kragh and Karstoft 87 ✓ ✓ ✓ ✗
Merican et al. 88 ✗ ✗ ✓ ✓
Raef and Ferdousi 4 ✗ ✗ ✓ ✓
Medenica et al. 89 ✓ ✓ ✗ ✓
Afnan et al. 9 ✓ ✓ ✗ ✗
Fernandez et al. 13 ✗ ✗ ✓ ✗
Curchoe 90 ✓ ✓ ✗ ✓
Tran et al. 91 ✓ ✓ ✓ ✗
Abdullah et al. 3 ✗ ✓ ✓ ✓
Current study ✓ ✓ ✓ ✓
Abbreviations: AI: Artificial intelligence; ART: Assisted reproductive technology.

Volume 2 Issue 3 (2025) 14 https://doi.org/10.36922/aih.7170