
Artificial intelligence is being used in IVF to select embryos with the highest chance of resulting in a successful pregnancy.
The AI algorithm, called Ivy, analyses time-lapse videos of embryos as they are incubated after being fertilised, and identifies which ones have the highest likelihood of successful development.
It was developed by Harrison.ai, a tech firm in Sydney, and has been used for several thousand women undergoing IVF in Australia. Women who undergo IVF using Ivy are informed about the algorithm and consent to its use.
Advertisement
Ivy was trained on more than 10,000 videos of embryos growing inside an incubator for five days, in combination with data about which embryos resulted in pregnancy. The model is more accurate than trained specialists at predicting the likelihood that an embryo will go on to develop a fetal heartbeat at or beyond seven weeks’ gestation.
In a , Ivy scored 0.93 on a measure known as AUC, where a score of 1 indicates a model with predictions that are 100 per cent correct. In comparison, trained human embryologists have previously scored an .
“The results look very promising,” says Iman Hajirasouliha at Cornell University in New York. He and colleagues have developed a similar AI, Stork, based on videos of 12,000 embryos on day five post-fertilisation.
What these AIs are identifying in the videos as a marker of success isn’t entirely clear. “Ultimately it’s still a bit of a black box,” says Aengus Tran, CEO of Harrison.ai.
“We could speculate that it must have learned very similar features to what embryologists are looking at,” says Tran. These might include the shape of cells, how symmetrically they divide, and how long it takes to go from two to eight cells.
Ivy may also have picked up on features that human specialists don’t look at, says Tran. While embryologists look only at video snippets of key events of an embryo’s development, Ivy is able to rapidly analyse the entirety of five days of footage.
Ivy is now routinely used as a tool in embryo selection by IVF Australia, which provides about 20 per cent of the country’s IVF services. “Because we think it’s a better predictor than the embryologists, we are choosing the embryos with Ivy most of the time,” says Peter Illingworth at IVF Australia.
A clinical trial will begin next month to determine whether women undergoing IVF whose embryos are selected by Ivy successfully become pregnant more quickly than women whose embryos are selected by human specialists.
In an average IVF cycle, a woman starts with 10 eggs that result in three potentially viable fertilised embryos after five days, says Illingworth. One of these is implanted, and the remainder are frozen and stored.
“Ultimately what you really want is one IVF cycle resulting in one embryo being transferred and one healthy baby,” says Tran.
“What we’re not doing is using Ivy to decide which embryos we will freeze and keep,” says Illingworth. “Ivy will never make the difference between a woman having a baby or not having a baby,” he says.
But it may – if the clinical trial shows the AI is successful – help a woman get pregnant more quickly by using the right embryo first.