Welcome back to Healthy Innovations! 👋

Fertility touches family, friends, and colleagues more than most of us realize – statistically, one in six people globally will face it at some point in their lives. That scale is exactly what makes the innovation happening in this space so exciting right now.

AI is entering IVF not with fanfare, but step by step: finding sperm that wasn't there, scoring embryos with a consistency no human eye can match, personalizing protocols that were once built for the average patient. The science is moving fast.

Let's dive in.

In a story published by TIME magazine in June 2025, we meet Rosie and her husband, who spent nearly 19 years trying to have a child. Across 15 IVF cycles, multiple surgeries, and specialists flown in from overseas, no one could find a single viable sperm. Her husband had azoospermia – a condition where no detectable sperm appears in the ejaculate – and by early 2025, every conventional path had closed.

Then they were referred to Columbia University Fertility Center, where a specialist offered one final option: an experimental AI system called STAR (Sperm Tracking and Recovery). STAR combines a custom microfluidic chip with high-speed imaging and a computer vision model trained to recognize sperm cells in what looks, under a standard microscope, like a sea of cellular debris.

In one of her husband's samples that trained embryologists had searched manually for two days without success, STAR found 44 viable sperm in under an hour. Rosie became pregnant in March 2025.

One pregnancy does not constitute a clinical revolution. But it captures something important about where fertility medicine is heading: AI is entering IVF as a series of targeted interventions across a process that has resisted meaningful improvement for decades.

IVF has been in clinical use since 1978 and the fundamentals have changed surprisingly little. Eggs are retrieved, fertilized in a laboratory, and a resulting embryo is selected for transfer, with embryologists making that final call based on how the embryo looks under a microscope.

Visual morphology is an imprecise predictor of which embryo will actually implant. On average, only about one-third of IVF cycles result in pregnancy, fewer lead to a live birth, and for women over 40 the numbers fall further still. Embryo selection is the highest-leverage point in the process – and an embryologist, however skilled, brings an inherently subjective eye, potentially scoring the same embryo differently on different days.

The most advanced AI embryo selection tools use time-lapse imaging – recording continuous footage of embryos developing over five to six days, then running those sequences through a deep learning model to produce a viability score. Vitrolife's iDAScore, trained on more than 180,000 embryo sequences with known outcomes, is the most studied system in this space. Its core advantage is consistency: the same embryo, scored at any time or in any clinic, returns the same result.

But the clinical evidence is more nuanced than the technology's reputation suggests. A 2024 randomized controlled trial published in Nature Medicine – the first of its kind – compared iDAScore against trained embryologists across 14 clinics in Australia and Europe.

Key results:

AI selection averaged 21 seconds versus over 200 seconds for manual assessment – a real efficiency gain. And the trial used an earlier algorithm version. What the result underlines is that AI embryo scoring is currently better understood as a consistency tool than a replacement for expert judgment. The two are most likely complementary.

Where AI is making a more clear-cut difference is in tasks that fall outside what human perception can reliably achieve at scale. STAR is the clearest example — finding viable cells in samples human experts had declared empty. But the same principle is driving innovation across the entire IVF process.

Automating the laboratory. In January 2026, Bloomberg Businessweek profiled AURA – a 17-foot, 4,500-pound robotic assembly line built by Conceivable Life Sciences, now operational at Hope IVF in Mexico City. AURA automates more than 200 steps involved in creating an embryo, from sperm preparation through to vitrification, using AI, advanced optics, and microrobotics to manipulate cells at a microscopic scale.

Personalizing stimulation. In ovarian stimulation – where medication dosing decisions determine how many mature eggs are retrieved – AI tools including Alife Health's Stim Assist draw on large historical datasets to suggest personalized protocols rather than applying population averages to individual patients. This matters because over-stimulation carries real clinical risk, and under-stimulation means fewer viable embryos.

AI embryo tools are advancing faster than the clinical evidence base supporting them. Most studies to date are retrospective, and the one landmark prospective RCT returned results more complicated than the field's enthusiasm had anticipated.

There are harder questions too. Current AI embryo selection tools score embryos on viability – the likelihood of implantation and progression to a heartbeat. They do not, and by regulatory design should not, select on any characteristic of the future child. That boundary is holding. The question is whether it continues to as commercial pressures grow.

Regulatory fragmentation is slowing adoption. The EU AI Act tightly governs high-risk medical AI; the FDA treats AI embryo grading as a medical device requiring premarket approval. Many promising applications remain in research settings only – not yet available in routine care.

The current generation of AI tools is working with data we already have. What comes next goes deeper.

Non-invasive embryo testing. The culture medium surrounding an embryo contains proteins and metabolites that may reveal viability markers invisible to any camera. AI analysis of this "spent medium" is still early-stage, but could eventually reduce or replace embryo biopsy – a significant step for both safety and cost.

Digital twins. Virtual models built from individual patient data could allow clinicians to simulate treatment responses before committing to a protocol – personalized planning before a single injection is given.

For Rosie and her husband, none of these future possibilities were what mattered. What mattered was 44 sperm found in an hour, in a sample declared empty after two days of human effort. By June 2025, she was five months pregnant and receiving standard obstetric care – the first confirmed pregnancy via the STAR system. "I still wake up in the morning and can't believe if this is true or not," she told TIME. After 19 years, that disbelief is the point.

🫟 Chemical staining has been the backbone of pathology for over a century. It works – but it's slow, consumes the sample, and quality varies between labs.

Pictor Labs is building the digital alternative. Founded by UCLA researchers, the company uses AI to generate stain-equivalent tissue images from unstained samples – no chemicals required. Their products cover virtual H&E, multiple stain types, and re-staining of existing slides.

Pictor positions itself squarely at the intersection of AI, histopathology, and lab automation – a space that is attracting serious attention as health systems look for ways to reduce bottlenecks in pathology workflows without sacrificing diagnostic quality.

🏥 When your hotel stay comes with an MRI. Sanford Health in Sioux Falls, South Dakota has done something the hospitality industry probably never saw coming: put a hotel inside a hospital.

The Highpoint Hotel occupies the top two floors of the newly opened Sanford Orthopedic Hospital – a nine-storey building that also houses 12 operating rooms, 19 inpatient rooms, and an intraoperative MRI. Patients facing a big procedure can now check in the night before via an elevator ride rather than a drive halfway across South Dakota, which in winter is less a commute and more a life choice.

The building was designed to make the single structure feel like two entirely different places. The hospital side is white-walled and clinical. The hotel side has wood accents, plush furnishings, a bar, a restaurant, and a sky lobby with a fireplace. The architects were so committed to the distinction that they specified different cleaning supplies for each floor – because nothing undermines a relaxing pre-op night's sleep quite like a hotel room that smells like a ward!

Thank you for reading the Healthy Innovations newsletter!

Keep an eye out for next week’s issue, where I will highlight the healthcare innovations you need to know about.

Have a great week!

Alison ✨

P.S. If you enjoyed reading the Healthy Innovations newsletter, please subscribe so I know the content is valuable to you!