Page 21 - AIH-2-3

Artificial Intelligence in Health AI in embryo selection for ART

(iv) Generative models for synthetic data: Images of requirement for open, and peer-reviewed research remain
synthetic embryos are produced using generative to be resolved. The application of AI as a quality control
models, such as diffusion models and generative tool post-thawing or for continuous monitoring of embryo
adversarial networks. Combining these artificial culturing systems optimizes laboratory workflows. This
images with actual data enhances AI model training application allows for a synergistic approach, integrating
and improves classification performance. 93 the strengths of AI algorithms with the expertise of
(v) AI-powered embryo ranking systems: Deep embryologists and healthcare professionals to improve
learning is used by platforms such as the Embryo ART outcomes. AI can streamline laboratory workflows
Ranking Intelligent Classification Algorithm to rank by automating time-consuming tasks, such as embryo
embryos according to their expected genetic state evaluation, allowing professionals to focus on critical
and implantation potential. These systems enable decision-making. 87
clinicians to select embryos with the highest potential
for successful implantation and pregnancy. 95 4.3.4. RQ4: What is the impact of AI-driven embryo
(vi) Automated morphological feature extraction: AI selection on the psychological health and decision-
solutions reduce subjectivity and unpredictability making processes of prospective parents, particularly
in assessments by automating the measurement of in light of ethical concerns?
important morphological parameters from embryo It is necessary for ethical considerations to maintain public
images. 96 trust and improve both psychological and clinical results
in ART. The use of AI technologies in embryo selection
101
4.3.2. RQ2: To what extent does ART improve the raises ethical concerns regarding deskilling, transparency,
pregnancy success rate and live birth outcomes accountability, and fairness. The absence of transparency
compared to traditional methods? in AI models, which are frequently referred to as “black-
The performance of ART may be improved using AI and box” systems, creates uncertainty and undermines trust.
98
ML approaches. These techniques also hold a promise for Furthermore, the potential repercussions of AI failures
the advancement of medical technology in the future. in embryo selection, which could lead to anomalies
5
AI-guided ARTs offer enhanced accuracy, uniformity, or undesired outcomes, underline the significance of
and automation compared to traditional ARTs. AI can responsibility and fairness in the utilization of AI in this
76
accurately identify embryos’ inner cell mass, blastocoel, sensitive field. To address these issues and preserve
71
trophectoderm, and zona pellucida. Deep learning the trust of the public, it is essential to prioritize the
methods are employed to accomplish this capability and development of AI models that can be interpreted, carry
reduce the workload of embryologists, thereby improving out thorough evaluations through randomized controlled
the efficiency of ARTs. In general, factors such as trials, and establish regulatory oversight for the application
97
maternal age, the underlying cause of infertility, and the of these algorithms. If these recommendations are
102
ovarian stimulation protocols significantly influence the followed, the integration of AI in embryo selection may
probability of achieving successful pregnancies and live enhance psychological and therapeutic results while
births. 22 upholding ethical standards in ART.
4.3.3. RQ3: How can AI algorithms be seamlessly 4.3.5. RQ5: What are the primary barriers and
integrated into existing embryo selection protocols limitations to the clinical application of AI algorithms
and laboratory workflows to leverage the expertise of in embryo selection, and how are these technologies
embryologists and healthcare professionals? being developed and validated?
AI algorithms can enhance embryo selection protocols AI-based ARTs can reduce inconsistencies, increase clinical
by leveraging data from time-lapse imaging, proteomic and client outcomes, and improve sperm testing and oocyte
profiles, and morphological features to predict live quality assessment. Obstacles to applying AI algorithms
103
birth outcomes and embryo viability. 73,79,98 Although arise due to the lack of transparency in ML models,
ML systems are used to predict the results of frozen ethical concerns about selection errors, and the fact that
ETs in early pregnancy, their accuracy is limited, and current embryo-selection algorithms lose diagnostic value
additional predictors are required to improve predictive when applied externally to many known implantation
performance. By utilizing ML models and computer embryos. These limitations are exacerbated by the black-
104
99
vision, AI can analyze vast amounts of image data to box nature of AI models, leading to ethical and epistemic
automate embryo selection processes. Nevertheless, issues, such as unclear responsibility for treatment success
100
issues such as the interpretability of AI models and the and biases with unintended consequences. To address
105
Volume 2 Issue 3 (2025) 15 https://doi.org/10.36922/aih.7170