World’s first baby born by IVF done almost entirely by a machine

A highly automated form of in vitro fertilisation (IVF) has led to a successful birth, raising hopes that this approach could cut the risk of human error during such procedures.

One method of IVF is intracytoplasmic sperm injection (ICSI), where sperm is injected into eggs in a lab dish. This is commonly used in cases of male infertility, as the sperm don’t have to work to reach an egg. Any resulting embryos are then inserted into the uterus. IVF can also be done by mixing sperm and eggs in a lab dish in the hope that fertilisation will take place, which is generally less successful, but also requires less medical intervention.

ICSI also has drawbacks, as it relies on high levels of precision and judgement from medical professionals. “They become tired and distracted sometimes, just like everybody else in most professions, so errors happen that can reduce [the] odds of fertilisation and births,” says Jacques Cohen at Conceivable Life Sciences, a biotech company in New York City.

To address this, Cohen and his colleagues have developed a machine that can perform 23 key steps required for ICSI. Each one is initiated by a person through the press of a button as they watch a livestream of the process. This can even be done from a different part of the world.

In one step, the machine uses an AI model to select the healthiest sperm cells for fertilisation, based on their appearance. In another, the machine immobilises the sperm by zapping their tails with a laser so they are easier to pick up. It later injects the sperm into already-collected eggs. A similar approach has been tested before, resulting in two live births, but some steps weren’t done by a machine.

To put their machine to the test, the researchers recruited a couple who were struggling to conceive, partly because the man had sperm that couldn’t swim properly. The woman also had problems producing eggs, so donor ones were used for the procedure.

The researchers randomly assigned five out of eight donor eggs to be fertilised by the automated system, which produced four embryos. The three remaining eggs were fertilised using the standard manual ICSI approach, all of which formed embryos.

They then used another AI model to select the two best embryos, based on the appearance of their chromosomes. Both of these were produced using the automated system, but that doesn’t necessarily mean this approach leads to healthier embryos than manual ICSI, says Cohen. We can’t gauge this due to the small number of eggs involved, he says.

When the team inserted one of the embryos into the woman’s uterus, it failed to develop, but the second led to a successful birth.

It is an exciting proof-of-concept, says Joyce Harper at University College London. But larger studies that randomly assign couples to undergo either the automated or the manual lCSI procedure are needed to establish whether the former approach leads to higher birth rates, she says.

Automated IVF is also unlikely to be widely used, due to it being an additional expense, at least when initially rolled out, says Harper. But Cohen hopes this would improve with time. “As we optimise, standardise and refine the system, we expect the cost to the patient and clinic to decline,” he says.