However, as AI becomes increasingly woven into the fabric of our lives, it brings with it a host of ethical concerns and considerations, raising questions about privacy, bias, and accountability. With so many aspects of AI that are truly beneficial to society and mankind, there’s no turning back time on it. Of course, in the wrong hands, AI can be terribly destructive, and that’s without talking about scaremongering and exaggerated claims of what AI can become if we let it. To maximise its benefits and mitigate its threats, we must create a framework for AI’s responsible development and application, remaining vigilant to ensure that it works for us, instead of allowing ourselves to work for AI.
This is particularly important in the realm of healthcare, where AI intersects with human life.
Within the realm of healthcare, AI has made traction in radiology and is beginning to interact with the very creation of life in reproductive care. Playing a pivotal role in offering renewed hope to aspiring parents, AI has the potential to transform the entire landscape of reproductive health analysis and intervention.
Consider IVF as a case study. Success rates for IVF treatments haven’t seen much improvement over the last few decades. In fact, only around 30% of women will realise a live birth in their first round of IVF. During an IVF cycle, a woman undergoes a regimen of hormones to stimulate the ovaries. She takes medication to help mature the eggs for collection, which takes place under sedation. A reproductive endocrinologist performs egg retrieval, extracting eggs, which will then be combined with sperm to form embryos – the earliest stages of life.
It is the role of the embryologist to care for the embryos, which are housed in incubators to mature. Embryologists, trained in the study of the formation, growth, and development of embryos, are tasked with assessing the quality of these and then selecting which to transfer in the hopes of that embryo implanting and leading to a live birth. They also determine which should be cryopreserved and which should be discarded. Though embryologists are extremely well educated and highly skilled, success rates for embryo selection are subject to human bias, experience, and expertise, leading to inconsistencies and overall varied and lower success rates than we would want.
Lacking a consistent system for collecting and interpreting end-to-end data on this process, IVF offers an ideal use case for AI to make a positive contribution.
Typically, embryologists place a fertilised egg in an incubator and manually check the progress under a microscope over a three-to-six-day period. The discretion is up to the embryologist on how many times over the process they assess and make notes on the embryos’ development, typically assessing on day one, day two or three, and then deciding on day four or five if the embryos are viable.
The process of carefully removing a developing embryo from an incubator to observe under a microscope and notate the data is time-consuming. Furthermore, removing embryos from the safe and constant environment of the incubator poses a risk of harm. The highly technical task of safely maintaining the beginnings of life outside of the body leaves little room for error. Embryologists assess a developing embryo using approximately five data points, known as the Gardner blastocyst grading system. This leaves IVF professionals to rely on their own education, basic assessment parameters, and their own experience of diagnosis.
In some practices, evaluating the embryo’s development is not done at all, instead waiting for the end of the process when a blastocyst is formed to check its quality and viability, and maybe send for genetic testing.
When assessing the embryo under a microscope, embryologists examine several morphologic characteristics to determine its viability. However, with the introduction of Time Lapse Incubators (TLIs) to the IVF lab, technologies, including machine vision and AI, can begin to have a greater impact. TLIs have advanced camera technology integrated into the incubator, capturing images of the developing embryos at regular intervals to create a “live stream”. While embryologists also have access to live stream images, they are humans like all of us, with limited capacity to absorb and analyse so much information. This is where AI outperforms humans. Having hundreds of images of developing embryos enables computer vision and AI technologies to assess the developmental stages of embryos based on millions of data points that are not obtainable by the human eye, to analyse that data, and even make predictions – objectively, consistently, and in seconds.
By training AI systems of large and diverse datasets that are representative of patients of varying ages, BMIs, races, and more, the AI can assist embryologists in assessing the viability of embryos. AI algorithms can automatically capture and process live data in TLIs, analysing the data in combination with additional patient data. Based on this, the AI provides additional insights and information to help embryologists make consistent, data-driven decisions. This, in turn, should result in more streamlined decision-making processes in clinics and among embryologists and lead to improved outcomes: pregnancy and live birth results.
AI calculates – it does not think. It calculates more effectively than humans: faster, more accurately, and more consistently, in order to bring new insights, so that humans can decide upon the best way forwards. Specifically, regarding both embryo and egg quality assessment, this ability to analyse vast, complex data maximises the potential success rates of IVF. The ability to analyse big data paves the way for improving human decision-making and the development of evidence-based parameters to establish a consistent standard of care.
AI enables a new era of precision embryo selection that streamlines the entire reproductive care journey for practitioners, augmenting operational efficiency and enabling a focus on personalised care, ultimately translating into superior patient outcomes. For patients, it can reduce the emotional and financial burdens associated with multiple IVF cycles and offer prospective parents a more engaged and transparent experience throughout the care journey.
Yet, this vision is only possible if AI is developed and implemented responsibly. And if clinicians use it wisely.
This brings us back to the initial question of responsible AI development and use, particularly when AI intersects with human life. It is not enough for AI to simply automate processes and provide a quality score. Rather, AI used in the field of reproductive care should qualify outcomes with explanations of how the AI arrived at a given conclusion, offering transparency on the process in quantifiable biological terms that practitioners understand.
Explainability and transparency form the cornerstone of ‘Responsible AI’ principles. It is only through transparent AI practices that we can strike a balance between technological innovation and ethical responsibility.
Beyond the specific domain of fertility, the requirement for transparency in AI development can shape the broader narrative surrounding AI’s responsible integration into healthcare and society at large. Transparency serves to demystify AI mechanisms, ensuring that algorithms are comprehensible to stakeholders. This commitment to transparency nurtures trust and engenders accountability of those who utilise this powerful technology.
Active participation in discussions, collaborations, and initiatives aimed at ensuring ethical and responsible AI use helps to advance judicious AI integration. As AI continues to weave its way into every aspect of human life, the ethical considerations surrounding its use become increasingly pressing. The realm of fertility serves as a poignant case study in balancing AI’s potential with ethical responsibility, illuminating the broader path forwards for responsible AI integration in society.
Eran Eshed is the CEO & co-founder of Fairtility, the transparent AI innovator powering in vitro fertilisation (IVF) for improved outcomes. He is a multidisciplinary business executive and serial entrepreneur with over 25 years of experience spanning numerous product and business domains. Eran was a co-founder and chief business officer of Altair Semiconductor, a wireless chipset innovator acquired by Sony in 2016. Eran holds a BsCEE in electronics engineering from Tel Aviv University.