BALTIMORE — An artificial intelligence (AI)-powered tool could be a last resort for couples wanting to conceive a biological child who had been told they have no sperm, researchers said.
The Sperm Tracking and Recovery (STAR) system detected sperm in all 10 spiked samples, while a clinician found sperm in only three samples, according to Hemant Suryawanshi, PhD, of Columbia University Fertility Center in New York, and colleagues.
Further, STAR detected a total of 167 sperm in those samples, compared with just four for manual detection, they reported at the American Society for Reproductive Medicine (ASRM) annual meeting.
Also, among 175 men with azoospermia, STAR detected at least one sperm in 26% of samples, and so far, STAR has led to at least three live births in which the recovered sperm were used for intracytoplasmic sperm injection (ICSI) or cryopreservation, Suryawanshi and colleagues reported.
“What this data shows is that a substantial number of these patients have rare sperm, and the limitation is not the biological absence of sperm for these cases, it’s actually our lack of ability to detect the sperm — and systems like STAR are trying to overcome that limitation,” Suryawanshi told MedPage Today.
Being male is a primary or contributing factor in about half of all infertility cases, the researchers said. About 10% to 15% of infertile men have azoospermia, though in actuality there may be so few it evades manual detection.
For couples wanting to conceive a biological child when the male has azoospermia, the next step is generally testicular sperm extraction (TESE) or microsurgical TESE, in which sperm is surgically extracted from the testes. This procedure can be painful and fails about half of the time.
Instead, Suryawanshi and colleagues developed the STAR tool to help detect ultra-rare sperm. It uses high-speed imaging, real-time AI, and microfluidic technology to take two million images, assess them, and isolate the sperm detected by AI into a small volume.
To assess whether STAR could reliably detect ultra-rare sperm, they spiked 40 sperm into an azoospermia sample, mixed it, split it in half, and gave half to an expert and half to STAR. Both were given 60 minutes to find sperm. This was done 10 times.
STAR found the sperm in all 10 cases while the manual expert only found it three times. Moreover, STAR found 167 sperm while the human only found four — a 41-fold difference.
Lastly, 175 consecutive patients were referred for STAR analysis. All of these men (median age 36) were deemed sperm-negative by manual examination before STAR. However, STAR detected at least one sperm in 26% of the samples. Moreover, at least five sperm were detected in 22% and at least 10 sperm were detected in 19%.
Some of these patients had undergone TESE and were still classified as sperm-negative; STAR detected sperm in 29% of this cohort.
Suryawanshi also highlighted that sperm identified with STAR has led to at least three live births in which the recovered sperm were used for ICSI or cryopreservation.
In one case, STAR found sperm in the ejaculate from a man who had underwent TESE twice. While the couple had a 19-year history of infertility, they successfully conceived and had a live birth in December 2025. This finding was reported as correspondence in The Lancet.
The other two live births occurred in July and this month, Suryawanshi reported.
He said a major limitation is the user-friendliness; STAR requires a highly-trained technician to run it, though his team is working on developing a simpler version and other groups are working on similar tools.
