Artificial intelligence in assisted reproductive technology: how best to optimize this tool of the future

Artificial intelligence, at a simple level, involves the use of a computer that can perform “human” functions: learning from experience, adjusting to new inputs, and simulating human intelligence performing human tasks. This Views and Reviews brings together a diverse group of investigators to evaluate artificial intelligence and the roles it might play in the field of assisted reproductive technology. Artificial intelligence, at a simple level, involves the use of a computer that can perform “human” functions: learning from experience, adjusting to new inputs, and simulating human intelligence performing human tasks. This Views and Reviews brings together a diverse group of investigators to evaluate artificial intelligence and the roles it might play in the field of assisted reproductive technology. The field of assisted reproductive technology (ART) is an example of how technologies, which develop from the slow, methodical pace of basic science, move rapidly into clinical care. Our young field is, as we know, barely 40 years old. Assisted reproductive technology is a classic example of “bench-to-bedside” medicine. We have been fortunate to have developed this field at a time of exponential growth in technology. This is critical for our complex field that not only encompasses basic physiology and endocrinology along with embryology and developmental biology but also requires precisely tuned machines that can assist us in the daily culture of human gametes and embryos. We are also witnessing an explosion of money in our field. Although largely focused on growth in clinical care, this input has the capacity to not only foster the growth of new technologies but also push us to accept the next “shiny thing” without thoughtful consideration of how best to incorporate the next new technology. We have seen this with many add-ons in both clinical care and the laboratory. Diagnostics and therapeutics have been introduced into practice without careful validation. At their worst, these purported “advances” add cost and time to the patient journey without any improvement in success–achieving a live birth. Artificial intelligence (AI) is the new “shiny thing.” However, this new “enabling” technology has the opportunity to transform how we practice. Already being incorporated into radiology (1Mello-Thoms C. Mello CAB Clinical applications of artificial intelligence in radiology.Br J Radiol. 2023; 96https://doi.org/10.1259/bjr.20221031Crossref Google Scholar), pathology (2Acs B. Rantalainen M. Hartman J. Artificial intelligence as the next step towards precision pathology. J.Intern Med. 2020; 288: 62-81Google Scholar), and cardiology (3Martínez-Sellés M. Marina-Breysse M. Current and future use of artificial intelligence in electrocardiography.J Cardiovasc Dev Dis. 2023; 10: 175-191Crossref PubMed Scopus (1) Google Scholar), the opportunities for ART are only now being explored aggressively. However, as with other fields, there are perils and pitfalls that we should not ignore as we become enamored with the promise of this new technology. With this in mind, we have brought together experts from within and outside of the ART world. Allowing a broader perspective is critical to our optimizing use of AI as its capacity grows and our awareness of new options for care develops. The lead article in this Views and Reviews by Brian Miloski (4Brown C. Nazeer R. Gibbs A. Le Page P. Mitchell A.R. Breaking bias: the role of artificial intelligence in improving clinical decision-making.Cureus. 2023; 15: e36415-e36428PubMed Google Scholar) introduces us to AI as a concept. He shares what AI is, what it can do, and what it cannot do. In this introduction, there is enthusiasm for the future but also caution about implementation. Drs. Hariton, Pavlovic, Fanton, and Jiang then dive into more specific applications of AI in ovarian stimulation as “computer-driven support systems.” They discuss current models, models in development, and future directions to improve the outcome with this most critical clinical aspect of ART. There are a number of key decisions in ovarian stimulation, including protocol selection, medication dosing, ultrasound and blood monitoring, and trigger time determination. Each of these lends itself to the analysis of big data and machine learning. In addition to discussing the aspects of ovarian stimulation that have been investigated, they also discuss the importance of validation—both internal and external. Drs. Jiang and Bormann take us into the embryology laboratory. The use of AI for embryo selection has received the most attention up to this point. As a subjective and skill-based task, it lends itself to an objective, trainable system for improving the efficiency and quality. However, although embryo selection is critically important, it is late in the embryology culture process. These investigators share how AI could improve many of the current subjective, largely manual processes within the laboratory, from the assessment of oocyte quality and sperm selection to fertilization and embryo culture and transfer. Although many of these new tools still require validation and testing, the potential to improve efficiency and quality along with increasing access is strong. The next section, by Cherouveim, et al., reviews approaches to improving sperm selection. The biology of sperm function, and how “selection” should be made when we overcome biological selection in the embryology lab, has received relatively little attention, despite the major role sperm plays in the reproductive process. The ability of AI to evaluate a large amount of data (thousands/millions of sperm) followed by pattern recognition and a learned approach may finally push andrology and subsequent embryology to new levels of success. Dr. Letterie then puts it all together and compares and contrasts the current state to the potential for a future with AI integration into practice. The image of greatest impact is the one characterized by systems that improve the clinical experience–for the patient and the practice. This is a tall order and starts from the early evaluation of the patient through clinical care/treatment and into the laboratory. Additionally, the importance of the publication of AI trials for the generation of good, evidence-based information on which to make decisions about usefulness is stressed. This is required of all new knowledge and equally applies to the incorporation of these new technologies. Lastly, Dr. Carol Lynn Burton Curchoe addresses the regulatory environment for AI tools. First, an overview of how standard approval for drugs and devices is handled by the Food and Drug Administration is provided. This is followed by a discussion of how AI technologies are viewed in the regulatory environment. Dr. Curchoe describes why the very aspect that makes AI appealing–its ability to learn and improve–increases complexity in the approval process. This section also addresses critical issues regarding cybersecurity and the ethics of these tools and their development. In summary, AI holds the potential to improve care for our patients and improve practice management by process streamlining and workflow optimization, thereby decreasing costs and improving access. This potential is currently driving research and advancements in the field. Additionally, AI tools (and they are tools—not replacements for human care) have the potential to eliminate bias in care and target the best care for individual patients. However, this is highly dependent on assuring quality input for these tools and the lack of automation bias (4Brown C. Nazeer R. Gibbs A. Le Page P. Mitchell A.R. Breaking bias: the role of artificial intelligence in improving clinical decision-making.Cureus. 2023; 15: e36415-e36428PubMed Google Scholar). It is not a matter of if AI will be integrated into our field, but rather how. This series challenges you to think broadly about the potential for AI in our field, but to do so with eyes wide open, so we anticipate the good and the concerning while striving to optimize utilization and improve patient care.