AI technology offers new hope for eradicating male infertility

A new technology based on artificial intelligence (AI) is now able to detect sperm cells even in men who were previously told they had no sperm in their bodies—and therefore were unable to father children. The breakthrough opens up new possibilities for couples who have been trying to have children for years.

Penelope was driving home from work in New Jersey, USA. It was early November 2025. Just then, a call came to her phone—the call she had been eagerly waiting for for many days.

Her doctor was on the other end of the phone. From him, Penelope learned that after two and a half years of painful trying, she had finally become pregnant.

After various tests, Penelope and her husband Samuel learned that Samuel had Klinefelter syndrome. It is a genetic condition in which men are born with an extra X chromosome and is often not diagnosed until adulthood.

Most men with Klinefelter syndrome have very few or no sperm in their semen. This condition is called azoospermia. About 10 percent of infertile men suffer from azoospermia.

Penelope waited with joy and disbelief until the evening when Samuel would return home to tell her the news.

“There was a wave of emotion on his face,” she says. “He cried… After all the hard work, time and research, we finally got to this point. We had just one embryo, and that was it. We were overjoyed to see it.”

Their pregnancy was made possible by a new technique called the STAR (Sperm Track and Recovery) system.

The method, developed by Columbia University, is used to find sperm in men with azoospermia. The system uses artificial intelligence to help identify and locate ‘hidden’ sperm, which can be present in small numbers in men with the condition.

‘I was terrified. I thought I wouldn’t be able to have children of my own, which is a huge part of my life,’ says Samuel.

He was told his chance of having a biological child was 20 percent.

‘That was a huge shock,’ he says.

Millions of men around the world suffer from infertility. One in six men of childbearing age will experience difficulty conceiving at some point in their lives.

In up to 50 percent of these cases, infertility is also a contributing factor, and one percent of men are azoospermic.

This means that there are probably millions of men around the world whose sperm counts are so low that it is very difficult to find any of their sperm, making them classified as azoospermic.

But the ability to find these hidden sperm with the power of AI could offer hope to aspiring parents.

After five years of development, the first baby using the STAR system was born late last year; a child was born to a couple who had struggled with infertility for nearly two decades.

The moment is still vivid in the minds of Jeff Williams, director of the Columbia University Fertility Center, and his team.

“Everyone was jumping for joy,” he says. “It’s rare that years of hard work can be rewarded with something so beautiful and special. Now there’s a baby girl, and hopefully, God willing, there will be many more.”

Since the birth of the first STAR baby, the technology has been used routinely at fertility centers, and the waiting list for people hoping to have children has grown to more than 100 people around the world.

Based on data from the last 175 patients, Williams said, they are finding sperm in about 30 percent of cases, even though they were told they had no chance of conceiving with their own sperm.

Further testing showed that the STAR system found 40 times more sperm than a manual search by a trained technician.

Typically, there are several hundred million sperm per milliliter of semen. The number of sperm is estimated by examining a tiny droplet from a sample under a microscope and seeing if they are motile and healthy.

But an azoospermic sample can contain just one sperm—or in some cases, none at all. It is not practical to sift through the entire sample drop by drop like this.

Williams got the idea for the STAR system after reading a report in 2020 that AI is being used to find new stars.

Modern telescopes generate vast amounts of data from the night sky, which can be very time-consuming for humans to analyze and identify unknown objects. But machine learning algorithms can do this in minutes.

“The skyscape we were looking at was very similar to what we see in men who are said to have no sperm,” Williams says.

He wondered if it was possible to use the same technology to identify and isolate sperm?

He and his team were already using a high-performance imaging technology that could scan the sample. The challenge was to analyze hundreds of images per second in real-time and identify and collect any sperm that were there.

Williams and his colleagues used microfluidic chips, made of glass or polymer with a series of channels as thin as a human hair. The sperm sample flows through the channels and is scanned by an imager.