Leveraging artificial intelligence and machine learning in assisted reproductive technology

INTRODUCTION

McCarthy, in 1955,^[1] coined the term artificial intelligence (AI). Since then, AI has developed rapidly and impacted our personal and social life. The last decade has witnessed the tsunami of computers. Driven by computer power, it has been possible to increase data storage, and with the enhanced memory, AI has been trained to handle increasingly complex tasks with unbelievable success. AI utilizes sophisticated algorithms to analyze vast quantities of biological data, uncovering potential relationships and leveraging these insights to augment clinical activities. Furthermore, by synthesizing knowledge from successful clinical cases and established guidelines, AI continuously enhances its accuracy. By mitigating diagnostic and treatment errors inherent in human clinical practice, AI enables real-time health risk predictions, thereby improving patient care outcomes. As technology continues to evolve, it is imperative to strike a balance between harnessing the potential benefits of AI and addressing the ethical considerations associated with its integration into reproductive medicine.^[2] Through collaborative efforts between healthcare professionals, researchers, and policymakers, the seamless integration of AI into assisted reproductive technology (ART) holds the promise of further improving outcomes and expanding access to fertility treatments for individuals and couples worldwide. AI plays a pivotal role in improving the accuracy and speed of diagnostic processes in reproductive medicine. Machine learning (ML) algorithms can analyze vast datasets, including medical histories, genetic information, and imaging results, to identify patterns and predict potential fertility issues.^[3,4]

AI refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. It encompasses learning, reasoning, and self-correction. At its core, AI involves the development of algorithms that enable machines to learn from data, allowing them to adapt to new inputs and perform human-like tasks with increasing accuracy. Initially conceptualized in the mid-20^th century, AI has evolved from theoretical research to practical applications, driven by advancements in computational power, data volume, and algorithmic complexity.^[5]

This article explores the multifaceted role of AI in the realm of ART, highlighting its contributions to diagnosis, treatment optimization, and the ethical considerations associated with its integration. AI plays a pivotal role in improving the accuracy and speed of diagnostic processes in reproductive medicine.^[6] ML algorithms can analyze vast datasets, including medical histories, genetic information, and imaging results.^[7] ML can also identify patterns and predict potential fertility issues. For instance, AI can assist in identifying specific genetic markers associated with infertility, allowing for more precise and personalized diagnoses. This enables healthcare professionals to tailor treatment plans based on individual patient profiles, leading to more effective and targeted interventions. Our objective is to provide a concise overview of AI methodologies, focusing on their relevance in the field of ART and medicine. This review delineates the potential constraints and obstacles while delving into the opportunities and future trajectories within the realm of in vitro fertilization (IVF).

UNDERSTANDING ART

The optimization of ART procedures is another area where AI demonstrates its transformative potential. IVF, a widely used ART technique, involves the selection of the most viable embryos for implantation. AI algorithms can analyze various factors, such as embryo morphology, genetic information, and patient history, to predict the likelihood of success for each embryo. This assists fertility specialists in making more informed decisions regarding embryo selection, ultimately increasing the chances of a successful pregnancy.

AI has been increasingly applied in assisted reproductive physiology over the past few decades. One of the earliest applications of AI in this field dates back to the 1990s when artificial neural networks were employed to analyze sperm characteristics and predict the success of IVF procedures.^[8]

TYPES OF ART PROCEDURES (IVF, INTRAUTERINE INSEMINATION [IUI], INTRACYTOPLASMIC SPERM INJECTION [ICSI], ETC.)

ART is not without challenges and limitations. ART has transformed the landscape of family-building, offering solutions and hope to individuals and couples facing fertility challenges. However, ethical, legal, and social implications surround the use of these technologies, emphasizing the importance of careful consideration and regulation in their application.

AI provides hope to couples who struggle with infertility due to various reasons, such as blocked fallopian tubes, low sperm count, or ovulation disorders. Techniques such as IVF, IUI, and ICSI can help these couples achieve pregnancies. The basic procedures of IVF are briefly depicted in Figure 1 and Table 1.

Close

Figure 1:

Schematic illustration showing the intracytoplasmic sperm injection procedure, where a single sperm is injected into the oocyte using a micropipette.

Export to PPT

AI IN REPRODUCTIVE CARE

AI is increasingly becoming a pivotal tool in enhancing reproductive health services, offering innovative solutions to complex challenges. By leveraging AI’s analytical power, healthcare providers can offer personalized, efficient, and more accurate reproductive care. IVF at present is a very subjective science, depending on the expertise and experience of the operators, mainly embryologists. Automation and AI are expected to bring about a more calculated, computed, and standardized approach to IVF. However, for many infertile patients, the journey to parenthood through IVF is a lengthy and emotionally and financially challenging process. The focus of a great deal of research is to improve on the current ~30% success rate of IVF.^[9,10]

For example, in dermatology AI tools, under-representation of darker skin tones in training images has resulted in lower diagnostic accuracy for melanoma in individuals with pigmented skin. Similarly, AI-based cardiovascular risk prediction models trained predominantly on data from Western populations have failed in predicting outcomes for certain ethnic groups, leading to potential under-treatment. In reproductive medicine, if AI embryo selection models are developed using datasets from specific demographic groups or laboratory conditions, they may not generalize well to diverse populations – potentially introducing biases in ART success rates. By incorporating these examples, we aim to highlight the importance of using diverse, representative datasets and pointing out the bias-mitigation strategies for AI model development in ART.^[11]

HISTORICAL CONTEXT

The integration of AI into reproductive care began with the development of predictive models and algorithms to improve IVF success rates. Early applications focused on analyzing embryo images to select the ones most likely to result in a successful pregnancy. ARTs, including procedures such as IVF, ICSI, and preimplantation genetic testing, have revolutionized fertility treatments. These methods address various infertility challenges by enabling conception outside the body and facilitating the selection of embryos with the highest potential for successful implantation.^[12]

Over the years, these technologies have evolved, incorporating more sophisticated ML models to analyze a broader range of data, including genetic information and patient health records.^[13] Since then, AI and ML techniques have been utilized in various aspects of ART.

A few concrete examples of different AI methodologies and their reported performance in clinical or laboratory settings.

1. Convolutional neural networks (CNNs) - Used in image-based embryo grading, CNN models such as STORK have achieved up to 97.5% accuracy in distinguishing high- and low-quality embryos, outperforming inter-observer agreement among embryologists.^[14]

2. Deep learning time-lapse models - Applied to predict implantation potential from sequential embryo development images, achieving an area under curve of 0.93 in some studies.^[15]

3. Multi-modal fusion models - Combining blastocyst morphology (through computer vision) with patient clinical data (through multilayer perceptrons) has shown improved predictive performance for clinical pregnancy (accuracy 82.4%) compared to either modality alone.^[16]

4. Decision support systems using random forest and gradient boosting - Employed for predicting ovarian stimulation outcomes and tailoring gonadotropin doses, leading to higher retrieval efficiency with reduced medication use.

This clarifies not only the AI techniques used (CNNs, deep learning, multi-modal fusion, and ML ensembles) but also the clinical or laboratory outcomes they achieved, thereby providing a clearer view of the scope and potential impact of these approaches in ART.

To assess the likelihood of IVF success, analyzing factors such as embryo quality, genetic data, and patient health indicators are to be considered.^[14] ML models, such as deep learning, have been instrumental in identifying patterns that may not be apparent to human clinicians. AI algorithms are employed in pre-implantation genetic screening and pre-implantation genetic diagnosis (PGD), helping identify embryos with the highest potential for healthy development by screening for genetic abnormalities. Natural language processing is used to scrutinize through vast amounts of unstructured data in patient health records, extracting relevant information for personalized treatment plans and fertility recommendations. AI-driven wearable technologies monitor vital signs and physiological parameters, providing real-time data that can be used to optimize fertility windows and improve reproductive health outcomes.

THE INTERSECTION OF AI AND ART

AI is instrumental in optimizing hormonal stimulation protocols in ART. By analyzing patient responses to medications and adjusting treatment plans in real-time, AI can enhance the efficiency of ovarian stimulation, minimizing side effects and improving overall treatment outcomes. This dynamic approach to treatment personalization showcases the potential for AI to revolutionize the entire ART process.^[10] While the integration of AI in ART brings about numerous benefits, it also raises ethical considerations that must be carefully addressed.

Leveraging sophisticated algorithms, AI tools are designed for personalized treatment plans that integrate a patient’s medical history, genetic profile, and responses to previous therapies. This precision-driven approach reduces trial-and-error in medication selection, supports real-time monitoring, and allows timely adjustments to optimize outcomes, ultimately enhancing patient care. When responsibly integrated into IVF stimulation protocols, AI has the potential to elevate clinical standards, making fertility treatments more effective, efficient, and accessible.^[17]

One significant concern is the potential for bias in AI algorithms, which may inadvertently perpetuate existing disparities in healthcare. Ensuring that AI models are trained on diverse datasets and continuously monitored for bias is crucial in maintaining equity in reproductive medicine. Another ethical consideration is the need for transparency and informed consent. Patients must be adequately informed about the use of AI in their treatment and understand the implications of relying on machine-driven decisions. Open communication between healthcare providers and patients is essential to foster trust and ensure that individuals undergoing ART feel empowered and well-informed throughout the process.

AI AND ML IN PRE-ART DIAGNOSTICS

The role of AI in ART is rapidly expanding, offering unprecedented advancements in diagnostic precision, treatment optimization, and overall efficiency.^[18] As technology continues to evolve, it is imperative to strike a balance between harnessing the potential benefits of AI and addressing the ethical considerations associated with its integration into reproductive medicine. Through collaborative efforts between healthcare professionals, researchers, and policymakers, the seamless integration of AI into ART holds the promise of further improving outcomes and expanding access to fertility treatments for individuals and couples worldwide. While ART has provided hope and solutions to millions, it is not without its challenges [Table 2].

THE CONVERGENCE OF AI AND ART: TRANSFORMING REPRODUCTIVE MEDICINE

AI has come to stay, it is useful to welcome it and learn to use it for transforming reproductive healthcare. AI represents a powerful technological wave, endowing machines with the ability to execute cognitive functions such as perception, reasoning, learning, and interaction. This transformative technology has swiftly integrated into various aspects of our lives, addressing business challenges through the convergence of three key advancements: Algorithmic innovation, vast datasets, and the proliferation of cost-effective computational power and capacity.^[12]

Integrating AI within ART marks a pivotal shift in reproductive medicine [Figure 2]. This synergy leverages AI’s advanced capabilities in data analysis, pattern recognition, and predictive modeling to significantly enhance patient screening, embryo selection, and treatment personalization processes. The fusion of AI and ART promises to improve clinical outcomes and revolutionize the patient care model in fertility treatments.

Close

Figure 2:

Schematic illustration depicting the integration of artificial intelligence with assisted reproductive technology, highlighting the role of digital tools in fertility treatments. ART: Assisted reproductive technology

Export to PPT

AI TECHNOLOGIES IN ART

Utilizes historical data to predict future outcomes, enhancing the decision-making process in patient treatment plans and embryo selection. Deep learning is a subset of ML, deep learning with neural networks mimics human brain function to analyze complex medical data, including high-resolution images of embryos, to identify those with the highest potential for successful implantation.^[15]

The potential of AI, particularly ML, in healthcare is evident in its capacity to predict highly individualized treatments for each patient. By analyzing unique patient characteristics, ML can optimize therapeutic outcomes. AI further empowers healthcare professionals by enabling therapy customization based on genetic profiles and predicted responses, while also facilitating real-time monitoring and treatment adjustments.^[19]

Neural networks are used for image analysis in various ART procedures such as sperm analysis, embryo selection, and monitoring follicle development. CNNs are commonly employed to analyze images captured through microscopy to assess the quality of sperm, embryos, and follicle development. Computer vision employs algorithms to process and analyze images, significantly improving the accuracy of embryo assessment and selection by identifying morphological features correlated with higher implantation success rates.

CURRENT APPLICATIONS OF AI IN ART

AI-enhanced imaging for embryo selection

Utilization of computer vision and deep learning algorithms is useful to analyze embryo imagery, significantly improving the selection process for implantation. Studies have shown that AI can predict embryo viability with higher accuracy than traditional methods, leading to improved IVF success rates [Figure 3].

Close

Figure 3:

Schematic illustration depicting embryo selection for implantation, assisted by artificial intelligence algorithms, highlighting the role of digital analysis in improving assisted reproductive technology outcomes. AI: Arrow indicates using artificial intelligence (AI) for selecting the perfect embryo

Export to PPT

IDENTIFYING THE CURRENT GAPS IN AI AND ML RESEARCH WITHIN ART

First, there is a limited generalizability of AI models, as many current embryo selection and treatment prediction algorithms are trained on small, single-center datasets that lack demographic and geographic diversity, reducing their applicability to broader patient populations. Second, although AI excels in image analysis, there is a lack of robust models integrating imaging with genetic, hormonal, and clinical data to produce comprehensive, personalized predictions. Third, there is no consensus on standardized performance metrics, making it difficult to compare results across studies or ensure reproducibility. Finally, regarding ethical concerns – insufficient guidance exists on addressing bias, preventing genetic discrimination, and ensuring transparency in AI-driven decision-making.^[16]

Few studies have reported large-scale, prospective clinical validation of AI tools in ART, but, limiting the translation of promising lab-based models into routine practice.

PRACTICAL APPLICATIONS IN REPRODUCTIVE CARE

AI functioning as a sophisticated tool assesses its current environment and strategically takes actions to enhance the likelihood of achieving its objectives. In the field of medicine, AI serves not as a replacement for clinical expertise, but rather as a facilitator and analyzer, aiding in clinical decision-making processes. It involves the integration of vast clinical datasets and the development of computer algorithms utilizing factors from ML and deep learning technologies. In the realm of ART, the potential for AI support is extensive [Table 3]. An AI-powered ART software offers numerous advantages, including reducing inter-observer variability, optimizing medication dosages for oocyte stimulation to minimize adverse effects such as hyper-stimulation, thereby enhancing both clinical outcomes and user efficiency. In addition, it enables better assessment of sperm samples, evaluation of oocyte quality, and selection of embryos, further improving the efficacy of ART [Figure 4].^[20]

Close

Figure 4:

Schematic illustration showing sperm motility analysis using machine learning tools, where sperm movement is captured and processed through artificial intelligence-driven computational interfaces. AI: Artificial intelligence.

Export to PPT

THE TANGIBLE BENEFITS OF AI IN ART

NURTURING THE FUTURE: THE SCIENTIFIC COMMUNITY’S ROLE IN AI AND ART, GUIDING ETHICAL, AND EFFECTIVE INTEGRATION

As we embrace the potential of AI in ART, the responsibility falls on the scientific community to steer this integration toward the most ethical and effective outcomes.^[23] Scientists, researchers, and clinicians are at the forefront of pioneering a future where AI not only enhances reproductive medicine but does so with the highest standards of equity, accessibility, and respect for human dignity.^[24]

ETHICAL RESEARCH AND ACCESSIBLE TECHNOLOGIES

It is imperative to ensure that research in AI and ART adheres to strict ethical guidelines.^[25,26] This involves transparent study designs, the protection of patient data, and the consideration of long-term societal impacts. When employing AI, multiple professionals participate in diagnosing and selecting the suitable strategy, prompting inquiries into ultimate accountability amidst errors.

Concerns emerge regarding resource distribution and cost/ benefit ratio. Therefore, before integrating AI into clinical settings, a thorough evaluation of both the quantity and quality of utilized data is imperative, along with addressing transparency issues. Furthermore, it becomes essential to confront the adverse effects and societal dynamics potentially associated with the utilization of AI in reproductive medicine.^[27] Developing technologies that are accessible to all segments of society are crucial. This includes focusing on reducing costs, simplifying treatment processes, and ensuring that advancements are not limited to high-resource settings.

Empowering healthcare providers includes understanding the capabilities, limitations, and ethical considerations of AI tools, ensuring that providers can offer informed, compassionate care. Encouraging collaboration across disciplines – combining insights from reproductive medicine, computer science, ethics, and law – will be key to addressing the multifaceted challenges and opportunities presented by AI in ART. These advancements have not only improved success rates but also made fertility treatments more accessible to a broader range of individuals and couples, including those with genetic concerns and same-sex couples. Scientists and researchers have a role in advocating for policies that foster innovation while ensuring patient safety and privacy. This includes engaging with policymakers to understand the potential and limitations of AI in ART.^[28,29] Thus, AI can be used to contribute to the development of regulatory frameworks that are adaptive to technological advancements, ensuring that regulations are both supportive of innovation and protective of individuals’ rights.

CONCLUSION

Most recently it has been seen that Artificial Intelligence in ART is most successfully used in image-based applications (sperm, oocyte, embryo selection) and in predictive modeling (success rates, stimulation protocols, endometrial receptivity). These applications are now routinely used in advanced IVF centers worldwide, often as decision-support systems rather than replacements for embryologists. The integration of AI into ART presents a unique opportunity to redefine the boundaries of what is possible in reproductive medicine, especially in IVF technology. As we navigate this promising yet complex landscape, the scientific community’s role in guiding this journey – through ethical research, interdisciplinary collaboration, and advocacy for supportive policies – will be crucial in ensuring that these technologies deliver the greatest societal benefit. Building on the promising results observed in embryo viability prediction and laboratory automation, it is now imperative to direct future research toward a few critical areas. (i) For bias mitigation and dataset diversity – there is a need for developing AI models trained on large, demographically diverse, and multi-center datasets to ensure reasonable performance across populations. (ii) Leveraging computer vision, genetic, hormonal, and clinical parameters in unified predictive models for truly personalized patient care. (iii) For ethical governance and regulation, it is essential to establish transparent guidelines to prevent genetic discrimination, safeguard data privacy, and promote algorithmic explainability. AI can continue to revolutionize ART – transforming embryo selection, improving implantation prediction, and optimizing treatment pathways – while upholding equity, safety, and patient trust. With scientific innovations, and AI’s transformative potential, this field can demonstrate tangible benefits for reproductive healthcare worldwide.