This is how an AI made an impossible pregnancy possible

For couples facing a diagnosis of male infertility severe, like azoospermia (the absence of sperm in the ejaculate), the path is usually bleak. The solution goes through different very expensive techniques and in many cases invasive that can become inaccessible and leave a couple without options to have a biological child. But science has found a way to provide a solution to this problem, and artificial intelligence is the protagonist.

The milestone. Columbia University has achieved the first clinical confirmation of a pregnancy using technology based on artificial intelligence to recover a sperm in a sample where there was none a priori. Now, this breakthrough offers hope for couples who are affected by severe male infertility.

This success was reported in an especially difficult case: a couple with a 19-year history of infertility and many failed attempts to conceive a child. In total, the couple faced 19 cycles of egg retrieval and two testicular removal procedures that were not producing any results. But finally AI achieved something that traditional methods could not.

Sperm hunter. The system used for this case is called STAR, which is the English acronym for Sperm Tracking and Recovery and it has three elements:

• A microfluidic chip: a small disposable plate through which the semen sample flows.
• High-speed imaging: A system that captures 300 frames per second as the sample progresses.
• A deep learning AI: the brain of the operation that will analyze all the images in search of the sperm that will give a new life along with the egg.

The system with all these elements has the capacity to analyze the sample at a rate of 400 microliters per hour, processing up to one million images during this time. AI, which is based on architecture “You Only Look Once“, scans each frame in real time in search of sperm candidates so that they can later be extracted from the sample.

But logically something so important cannot be entrusted to a single frame. That is why the AI, to be sure that it has seen a sperm, will use at least three frames where it is seen before giving the good news.

Although there is one detail to resolve: a sperm is like a needle in a haystack (or even worse). This means that the moment there is a confirmed positive there is no time for a human to ‘catch’ it, but rather the system has a microfluidic mechanism that is activated and isolates the specific sperm in a volume of 300 nanoliters that can now be used by an embryologist.

A success story. The couple in the studio was a major challenge. The man, 39, had azoospermia and had already gone through multiple manual searches and failed surgical extractions. The 37-year-old woman had a very low ovarian reserve.

At first, embryologists began to look for some type of sperm in the semen, but there were no good results. That is when STAR was chosen to analyze 2.5 million images in approximately 2 hours, giving a result of seven sperm and two of them being mobile.

Those two sperm, which would have gone completely unnoticed by the human eye (unless one was very lucky), were recovered and injected into two eggs. Both eggs fertilized and developed into embryos, which were transferred. Thirteen days later the good news arrived, as the patient had her first positive pregnancy test.

A new path. As we mentioned at the beginning, this new AI-based identification system gives a lot of hope to couples who may have many problems getting fertilized due to problems finding any sperm in the ejaculate. Something that is currently in the research phase and is in the United States, so its landing in Europe may take some time.

In Xataka | Having many children sounds great as a way to preserve the species. Until you start passing genetic mutations