The latest scientific advances in the world of IVF have shown how the application of Artificial Intelligence (AI) techniques can also revolutionize the world of assisted reproduction.
An AI trained with hundreds of embryo images is able to predict which of them will thrive with an unprecedented accuracy rate of 85%. This results not only in a reduction in the risk of harming the embryo by making microscopic inspections unnecessary, but also in an increase in the percentage of women who will be able to become mothers thanks to these treatments.
On October 10, the team of researchers from Imperial College London, and Cornell University in the USA, led by Nikita Zaninovic, presented their results to the American Society for Reproductive Medicine, and they have joined the race to patent algorithms and enter the lucartivo market of AI applications, valued in $ 7.5 billion.
Figure 1, AI can also increase the number of women fulfilling their desire to become mothers through IVF
His works explain how this technology can greatly reduce the risk of complications, caused in 70% of the cases to abnormalities in the embryo. Currently, doctors implant multiple embryos in order to increase the chances of success. In return, this technique also increases the risk of premature births, preclamsia, and other costly and painful complications in childbirth.
How does it work?
Embryology specialists should choose which blasts to implant, based on their appearance and regular growth rate. Studies suggest that if an AI is trained to learn to distinguish what is an aspect and a normal growth rate, it is feasible to achieve much more reliable results than those obtained by current methods, based on an eye inspection by specialists.
When human experts do this work, differences in classification methods can be seen from one laboratory to another, or even from one expert to another. Using more than 50,000 images to train the algorithm, the researchers obtained a success rate in embryo categorization of 97%. The algorithm also correctly selected which embryos were most likely to thrive and become healthy babies 85% of the time.
Does this mean that human specialists will no longer be needed?
According to an MIT study published in 2017, when comparing the performance of human equipment and robots, the most efficient result was that obtained by combined teams of humans and machines, being 85% higher than the rest of the teams. There has already been much talk about the great potential of human intelligence – artificial intelligence collaboration, and how AI gives us “superpowers” that allow us to overcome all kinds of human limitations.
Rather than making human workers unnecessary, what will change is the type of work that human workers will do. For example, if these types of techniques are implemented, the demand for IT support profiles will increase.
Figure 2, Embryologists classify embryos by eye inspection
The advantages of this technology go beyond increasing treatment success rates. They also make it possible to avoid great suffering for families. Approximately half of the abortions suffered by women in IVF treatments are caused by an abnormal number of chromosomes in the embryo. Therefore, being able to select the best embryos to implant is the key to the success of in vitro fertilization.
In the words of Dr. S. Zev Williams of Columbia University, also a member of Cornell’s laboratory who developed one of these techniques “ ” Infertility and abortions have been health problems since Ancient Times. It is even poetic to solve them using the most modern technologies”
If we take into account that more than 200,000 couples start IVF treatments each year, and that around 2/3 of them experience at least one cycle failure, it is easy to realize the scale and scope of this revolution.
As we have seen, the potential of Artificial Intelligence goes far beyond data analytics and robotics. The tandem artificial intelligence – human intelligence applied to the world of health has a clearly promising future.
Translation of the original post written by Lucy Beardsley.
