• Integration of artificial intelligence with genetic and epigenetic data to predict implantation success and reduce the likelihood of miscarriage in the IVF process
  • Asal Abooei Mehrizi,1 Armin Motevalli Jooybari,2 Hiva Danesh Hosseini,3,*
    1. Department of Medical laboratory Sciences , TeMS.C., Islamic Azad University, Tehran , Iran
    2. Department of Medical laboratory Sciences , TeMS.C., Islamic Azad University, Tehran , Iran
    3. Department of Medical laboratory Sciences , TeMS.C., Islamic Azad University, Tehran , Iran


  • Introduction: In vitro fertilization (IVF) has transformed reproductive medicine, yet its success rates remain limited, with only 30–40% of cycles leading to live births. Implantation failure and early miscarriage are significant challenges, emphasizing the need for improved predictive strategies. The implantation process is influenced by complex factors such as embryo genetic integrity, endometrial receptivity, maternal age, and overall reproductive health. Conventional indicators like ovarian reserve and embryo morphology often fall short in explaining variability in IVF outcomes, prompting interest in genomic and epigenomic markers for deeper insights. Artificial intelligence (AI), particularly through machine learning and deep learning, offers promising advancements in this domain. By integrating diverse datasets—including imaging, clinical metrics, and multi-omics data—AI can enhance embryo selection, predict implantation success, and reduce miscarriage risks with greater precision. This review explores the intersection of AI and molecular biomarkers to address the biological complexities of IVF, highlighting recent progress, challenges, and future directions. Through personalized and data-driven approaches, AI stands to optimize fertility care, offering hope for improved outcomes in assisted reproductive technologies.
  • Methods: We conducted a systematic literature search in PubMed, Scopus, and Google Scholar databases up to June 2025. Predefined keywords related to artificial intelligence, in vitro fertilization, genomics, and epigenetics were used to identify relevant studies. Titles and abstracts were screened according to established inclusion and exclusion criteria. We included peer-reviewed original research articles investigating the application of AI techniques for predicting implantation success or miscarriage risk in IVF. Editorials, conference abstracts, and studies lacking methodological detail were excluded. Eligible studies underwent qualitative synthesis focusing on AI methodologies, datasets used, predictive outcomes, and reported challenges.
  • Results: The application of AI integrated with genetic and epigenetic data has led to notable advancements in reproductive medicine, particularly in improving IVF outcomes. AI techniques such as machine learning, deep learning, artificial neural networks (ANNs), convolutional neural networks (CNNs), and transformer-based models facilitate the analysis of complex, multidimensional datasets. These datasets include genetic variants, DNA methylation profiles, transcriptomic signatures, and imaging data. AI models have demonstrated enhanced capability to predict implantation success and miscarriage risk by capturing subtle biological signals that traditional methods miss. Moreover, AI-driven clinical decision support systems (CDSS) assist clinicians in embryo selection, optimize ovarian stimulation protocols, and automate laboratory workflows, thereby reducing human error and inter-observer variability. However, the field faces challenges including limited access to large, well-annotated datasets, demographic heterogeneity, and variability in data acquisition protocols that limit model generalizability. Ethical concerns such as the opacity of “black-box” AI models, potential biases, and patient trust remain significant barriers. Additionally, regulatory and legal frameworks addressing data privacy, liability, and clinical application of AI in IVF are still evolving, with compliance to regulations like the General Data Protection Regulation (GDPR) being mandatory.
  • Conclusion: The integration of AI with genomic and epigenomic data represents a promising paradigm shift towards personalized reproductive medicine. Tailored AI models can potentially improve prediction accuracy for implantation outcomes and miscarriage risks, enabling more informed clinical decisions. However, translation into routine clinical practice necessitates addressing current limitations. Ensuring transparency and interpretability of AI models is essential to foster clinician and patient trust. Developing robust AI systems requires collaborative efforts between reproductive specialists, data scientists, bioethicists, and policymakers. Future research should prioritize building diverse, high-quality datasets, advancing explainable AI techniques, and conducting rigorous prospective validation studies to evaluate real-world efficacy and safety. As these challenges are overcome, AI-driven tools are expected to become integral components of IVF workflows, ultimately improving success rates and patient care.
  • Keywords: Artificial Intelligence , In Vitro Fertilization , Endometrial Receptivity , Epigenetic Data