AI and IVF: A fertility doctor’s insight into the future of reproductive medicine

If it takes a village to raise a child, it takes a medical army to create one for infertile couples.

Today, global infertility rates are on the rise, but scaling up in vitro fertilization (IVF) services is no easy task. A single IVF cycle can be a lengthy process. Patients take a cocktail of hormones to stimulate their ovaries to create more eggs. These eggs are retrieved by a doctor and an embryologist then fertilizes the eggs with sperm in a petri dish. The number of eggs that become viable embryos decreases as they progress through the various stages of IVF. After several days, the doctor places all viable embryos back in the patient’s uterus. Each IVF round involves three to six visits to a clinic, with blood samples and ultrasounds to monitor the patient’s body to ensure that every step is being done at the right time. IVF is also expensive: Each round can cost between $15,000 to $30,000.

Dr. Amber Cooper is the chief medical officer of genomics and lab operation for the fertility and wellness clinic, Kindbody. Founded in 2018 by Gina Bartasi, Kindbody, now valued at $1.8 billion, offers fertility benefits directly to employers as well as owns and operates its own clinics. Dr. Cooper believes automation and AI can bring IVF to more people. Fast Company chatted with Dr. Cooper to understand what that would look like. The conversation has been edited for length and clarity.

Why is it important to increase access [to IVF]?

We are seeing increasing rates of infertility or subfertility: It used to be one in 10 couples, one in six couples. Now, we are almost at one in five couples. We need at least 10 times the access to IVF care that we have now. It’s not just infertility patients who need IVF. It is people with recurrent miscarriages or carrying a genetic disease and the LGBTQ+ population.

So how do we increase access?

We’re not suddenly going to have 10 times the number of doctors, 10 times the number of embryologists, and 10 times the number of clinics overnight. We need automation and AI. By automation, I mean making a process or system operate automatically so you replace human labor. And then you have AI, which is the ability to replace some human decision-making.

AI can automate manual steps related to blood processing and results, electronic medical record, billing, and other computer-mediated processes resulting in faster turnaround times and efficiency, which in turn may improve patient satisfaction and lower costs.

What is the status quo right now for IVF?

Most clinics and labs are using human processes and are not using automation or AI. The doctor does the procedure and retrieves the eggs. You have an andrologist who preps the sperm. And then you have your embryologist in the lab who injects the sperm into the egg to create an embryo. Then the doctor puts the embryo back in the body.

What does adding AI and automation to the IVF process look like?

Let’s break it down into each step of the patient journey. Probably eight out of a hundred people who need to see a reproductive endocrinologist ever get to our doors. First, there’s just getting people in the door and using AI models on social media to market to the right person.

Then there’s the possibility of using AI to predict the doses of medications to give people. There’s AI computer-driven programs already in the ultrasound space that you can use to monitor patients and the development of their eggs. There’s even technology that lets the patient do at-home blood testing and ultrasounds and send the info back to the clinic.

However, the really hot topic in AI is in the lab. So, to prep sperm, you take a needle to pick individual sperm and inject it into the egg to fertilize it. The fertilized eggs get moved to incubators to grow for several more days. This could all be done automatically with robots and AI, which some companies are already doing.

The other big place that AI is coming into is the inventory-tracking space. Right now, most clinics use a process called double witnessing, where two people confirm the right sticker is on the right tube. Adding an AI witness system adds protections for patients. More and more clinics are using RFID chips to help track dishes for sperm and eggs, chain of custody, etc.

What are the barriers to implementation?

A lot of embryologists think this will replace jobs. That’s not true. We need them on the more complicated steps. There are a lot of steps that can be simplified right now such as getting the media ready for the cells’ petri dishes. If we save embryologists time, we free them up for more complex cases, such as women who have fragile eggs, or when immature testicular sperm needs to be injected into an egg instead of mature ejaculated sperm, or when we have to biopsy embryos for complicated genetic diseases.

There’s a risk of decreased human interaction. In the patient’s eyes, you want to know someone is thinking about your case, the emotional context. On the business side, cost is a concern. How much will the robots and tech cost?

What does the future hold?

There’s the question of how good the AI models can get. Right now, to predict a woman’s response to medication, I might use the results of her blood test, the ultrasound count of follicles, her age and her BMI. AI might be able to create an algorithm to help. But there are always outliers who respond differently, which is where the art of medicine comes in. We’re trying to figure out how to incorporate that into AI models as well.

In addition, there’s the question of how far we can go with genetic testing. Currently, there’s a few basic areas where we do genetic testing. One is preimplantation genetic testing, which is looking to see if there’s too little DNA in certain chromosomes. That will result in an embryo that generally doesn’t implant. The second area is preimplantation genetic testing for a single gene mutation, such as cystic fibrosis or Huntington’s disease.

But there’s a new world emerging for polygenetic testing. Those are diseases that are not tied to a single gene. Instead, certain attributes of a genome might make someone more likely to have these diseases—think heart disease, autism, type 1 diabetes. There are companies that are offering testing for some of these polygenic diseases. We’re not close to a world where people can pick hair color or eye color, but we could presumably reduce disease burden, which is complicated.

AI is going to bring us to a point where there’s some really hard questions that need to be asked about the amount of human choice and intervention in terms of birth.