Scientists report the first ever clinical pregnancy aided by AI-guided sperm recovery. (Photo by Suhyeon Choi on Unsplash)

In A Nutshell

• A case report describes the first pregnancy using sperm detected and recovered by an AI-guided microfluidic system.
• The STAR chip scanned an azoospermic sample in real time and found seven sperm; two motile cells led to two embryos.
• Day-3 transfer resulted in a confirmed clinical pregnancy; an 8-week scan showed a normal heartbeat.
• It’s one case: larger studies are underway to see where this tool helps most and how clinics could use it.

After 19 years of trying to conceive, a couple facing severe male infertility achieved their first pregnancy using an artificial intelligence system that detected sperm invisible to human examination. Reported in The Lancet, this represents, to the authors’ knowledge, the first reported clinical pregnancy resulting from AI-guided sperm recovery. It offers new hope for men whose semen appears empty on manual examination, a diagnosis where traditional methods frequently fail.

A 39-year-old man had been diagnosed with non-obstructive azoospermia, meaning sperm are absent or extremely rare in his semen. His 37-year-old partner had undergone 19 egg retrievals across four fertility centers over 11 years. Despite invasive testicular procedures and extensive manual searches, the couple had produced only one embryo suitable for transfer in nearly two decades. That attempt failed.

When doctors examined the man’s semen sample under a microscope, they found nothing. But when researchers processed the same sample through the STAR system (Sperm Tracking and Recovery), the AI identified seven sperm after analyzing 2.5 million images in about two hours.

Two of those sperm could move on their own. Both were injected directly into eggs. Both eggs became embryos. Thirteen days after transfer, the woman received her first positive pregnancy test in 19 years.

How the System Works

The STAR system combines high-speed imaging with AI trained to recognize sperm. As the sample flows through a specialized chip, a camera captures 300 frames per second, analyzing over a million images per hour.

Rather than relying on a single snapshot, the system tracks each potential sperm across roughly ten consecutive frames. It confirms detection only when the object appears as sperm in at least three frames, which helps eliminate false positives from debris or other cells.

Once confirmed, a microfluidic gate isolates the sperm for retrieval. To validate accuracy, researchers tested the system by adding known numbers of sperm to samples. The results showed excellent correlation: R²=0.99.

Current options for severe male infertility can be both uncomfortable and tedious, all while hardly guaranteeing success. (Credit: Syda Productions on Shutterstock)

The system is fully enclosed to prevent contamination, and all components are single-use to ensure sterility. Both partners provided informed consent, and the research received ethical approval from Columbia University’s institutional review board.

When Standard Treatments Fall Short

Male factor infertility contributes to roughly 40 percent of all cases. Azoospermia and cryptozoospermia (extremely rare sperm) account for about 10 to 15 percent of male diagnoses.

Standard approaches include testicular sperm extraction, where surgeons remove small pieces of testicular tissue hoping to find sperm, or extended manual searches by embryologists. Both methods are invasive or time-intensive and frequently unsuccessful. Many couples are ultimately advised to consider sperm donors or adoption.

This couple had tried everything. The male partner underwent two testicular extractions with only rare sperm recovered. His partner’s hormone levels showed her egg supply was critically depleted. Her 19 egg retrievals produced consistently poor fertilization when sperm was available. Several cycles involved only egg freezing because sperm couldn’t be obtained.

The STAR system found seven sperm in a sample that appeared empty under manual examination. Small numbers, but enough for a chance.

The Results

Two of the seven sperm could move. These were injected into two eggs: one fresh, one thawed. Both became embryos.

Two additional thawed eggs were injected with non-moving sperm. Those didn’t progress. Only the motile sperm produced viable embryos.

Both embryos were transferred. Thirteen days later: positive pregnancy test. The couple’s first in nearly two decades. At eight weeks, ultrasound confirmed normal development with a heartbeat of 172 beats per minute.

While the STAR system had previously recovered sperm in other patients, this marks the first time AI-detected sperm were used immediately for fertilization and resulted in a confirmed pregnancy. That matters because it shows the system can support real-time clinical decisions, not just sperm banking.

The AI Behind the Detection

The AI model was built on the You Only Look Once architecture, trained specifically on sperm images. It divides each video frame into a grid and predicts where sperm might be in a single pass.

The model flagged sperm correctly 89 percent of the time and successfully identified 90 percent of actual sperm present. Processing at this speed required specialized computing power that would be impractical for manual examination. Even highly trained embryologists can examine only a fraction of a sample’s volume and may still miss rare sperm.

What This Means for Infertility Treatment

For men with severe male factor infertility, standard options can be uncomfortable or time-consuming without guaranteeing results. Many couples spend years cycling through repeated procedures, accumulating both financial costs and emotional exhaustion.

By automating detection and analyzing the entire sample volume, the STAR system creates opportunities where traditional methods have failed. It requires only a standard semen sample rather than surgical extraction, making it less invasive.

Larger clinical studies are underway to evaluate the system’s performance across broader patient populations. If those trials confirm consistent success, the technology could become a valuable addition to fertility clinics.

The technology may also help men with cryptozoospermia, where sperm are present but extremely rare, by enabling faster and more reliable detection.

Looking Ahead

At publication, the pregnancy was progressing normally at eight weeks. For this couple, who had exhausted nearly every option over 19 years, the system delivered what traditional approaches could not: a confirmed pregnancy using the male partner’s own sperm.

Whether they’ll ultimately deliver a healthy baby remains unknown. But they’ve already achieved something that seemed impossible just months earlier. And for thousands of other couples facing similar diagnoses, that represents a genuine reason for hope.

---

Disclaimer: This article discusses emerging fertility technology and research findings. It is intended for general information only and should not be considered medical advice. Please consult with a qualified healthcare provider for guidance on fertility treatment options.

---

Paper Summary

Methodology

The STAR system integrates high-speed imaging, custom microfluidics, and a deep learning model trained on annotated sperm images. The system captures phase-contrast images at 300 frames per second as samples flow through a specialized chip at 400 microliters per hour. The AI model, built on the You Only Look Once architecture, processes each frame in real time. It tracks objects across approximately ten consecutive frames and confirms detection only when an object appears as sperm in at least three frames. Once detected, a microfluidic mechanism isolates the sperm into a 300-nanoliter volume. Researchers validated accuracy by adding known quantities of sperm (5 to 100 per 400 microliters) to azoospermic samples and comparing detection counts to actual inputs.

Results

A 3.5-milliliter semen sample from a 39-year-old man with non-obstructive azoospermia was processed through the STAR system. Manual examination found no sperm. The STAR system analyzed approximately 2.5 million images over two hours and detected seven sperm: two motile and five non-motile. The two motile sperm were used for intracytoplasmic sperm injection into two mature eggs, both of which developed into cleavage-stage embryos. Two additional eggs injected with non-motile sperm did not progress. Both viable embryos were transferred on day three, resulting in a positive pregnancy test 13 days post-transfer. At eight weeks, ultrasound confirmed normal fetal development with a heartbeat of 172 beats per minute. In validation testing, the system demonstrated R²=0.99 correlation with spiked sperm counts. The AI model showed precision of 0.89, recall of 0.90, and mean average precision scores of 0.95.

Limitations

The report describes a single clinical case, which limits conclusions about reproducibility and efficacy across broader populations. The study does not provide success rate data across multiple patients or varied clinical presentations of azoospermia. The couple’s specific circumstances may not be representative of all couples facing similar diagnoses. Only two of seven detected sperm were motile, and only these resulted in viable embryos, raising questions about the clinical utility of detecting non-motile sperm. Long-term pregnancy outcomes, including delivery and neonatal health, were not available at publication. The study does not address cost, scalability, or practical implementation considerations.

Funding and Disclosures

The research was supported by the Havens Innovation Fund. Two authors, Hemant Suryawanshi and Zev Williams, are inventors on patent applications filed by Columbia University related to the STAR technology. All other authors declared no competing interests. De-identified technical and imaging data are available upon reasonable request, subject to institutional review and patient privacy regulations. Individual-level clinical data are not publicly available due to the single-case nature and patient privacy considerations.

Publication Details

Suryawanshi H, Gemmell LC, Morgan S, Koustas G, Prosser RW, Fu R, Forman EJ, Williams Z. First clinical pregnancy following AI-based microfluidic sperm detection and recovery in non-obstructive azoospermia. The Lancet. Published online October 31, 2025. DOI: 10.1016/S0140-6736(25)01623-X.

About StudyFinds Analysis

Called “brilliant,” “fantastic,” and “spot on” by scientists and researchers, our acclaimed StudyFinds Analysis articles are created using an exclusive AI-based model with complete human oversight by the StudyFinds Editorial Team. For these articles, we use an unparalleled LLM process across multiple systems to analyze entire journal papers, extract data, and create accurate, accessible content. Our writing and editing team proofreads and polishes each and every article before publishing. With recent studies showing that artificial intelligence can interpret scientific research as well as (or even better) than field experts and specialists, StudyFinds was among the earliest to adopt and test this technology before approving its widespread use on our site. We stand by our practice and continuously update our processes to ensure the very highest level of accuracy. Read our AI Policy (link below) for more information.

Our Editorial Process

StudyFinds publishes digestible, agenda-free, transparent research summaries that are intended to inform the reader as well as stir civil, educated debate. We do not agree nor disagree with any of the studies we post, rather, we encourage our readers to debate the veracity of the findings themselves. All articles published on StudyFinds are vetted by our editors prior to publication and include links back to the source or corresponding journal article, if possible.

Our Editorial Team

Steve Fink

Editor-in-Chief

Sophia Naughton

Associate Editor