We might imagine an IVF process to run in a straight line: eggs collected, sperm added, embryo transferred. In practice the process almost always stops partway and waits. Eggs are frozen after retrieval, sometimes for years, by patients preserving fertility ahead of treatment (like cancer) or before they are ready to use them. Embryos are frozen after five or six days of culture, so that a cycle producing several embryos can supply several attempts at pregnancy rather than one.

Whether the cell being preserved is an egg or an embryo, the procedure is the same. An embryologist loads it onto a plastic strip 0.4 mm wide, in a drop of about a tenth of a microliter, and plunges it into liquid nitrogen at –196°C. The cell has been sitting in cryoprotectant, a concentrated solution that protects it by replacing the water inside it and turns toxic if it stays there too long. The plunge has to be fast enough that the water left behind hardens into a glass rather than crystallising into ice. Automation delivers that plunge faster than an arm can move, and that speed buys the margin to shorten every step before it.

The Method

If you cool a living cell slowly, water inside and around it freezes into ice crystals. Those crystals act like microscopic shards of glass: they physically puncture and tear the cell’s outer membrane and its internal machinery. Eggs (oocytes) are especially vulnerable because they are large and water-rich. This is the core problem all freezing methods must solve.

The older approach, slow freezing, tried to manage ice rather than eliminate it. Using a programmable machine, cells were cooled very gradually, typically at rates of just –0.1 to –3°C per minute and the whole process took a minimum of about 90 minutes. It worked reasonably well for embryos but poorly for eggs, and results were inconsistent.

Vitrification takes the opposite strategy: go so fast that ice never has time to form. The word comes from the Latin for glass. Cells are bathed briefly in high concentrations of cryoprotectant, protective compounds that replace the water inside the cell, then plunged into liquid nitrogen at –196°C quickly enough that the liquid inside and around them hardens into a glass-like solid instead of crystallising. Rapid cooling is defined as a temperature reduction exceeding –2,500°C/min ( thousands of times faster than slow freezing). Modern minimal-volume devices go well beyond that threshold, reaching roughly –10,000 to –50,000°C/min and carrying the cell from room temperature to –196°C in under two seconds. Cooling rates of around 69,000°C/min, with warming at 118,000°C/min, have been measured using Cryotop carriers.

Vitrified embryos survive and implant well, enabling successive transfer rounds that make the most of the embryos generated in a single stimulation cycle; modern clinics also increasingly delay transfer to a later cycle (”freeze-all”), as the strategy can actually improve outcomes compared to fresh transfer. ESHRE’s European IVF Monitoring (EIM) Consortium1, covering 1,473 centres in 36 countries, recorded 401,483 frozen embryo transfer cycles versus 158,649 conventional (fresh) IVF cycles. Egg-freezing volume in the US rose from 24,560 in 2021 to 29,803 in 2022 and to nearly 40,000 in 2023. Vitrification is now the dominant cryopreservation method in IVF laboratories worldwide.

A landmark systematic review in 2017 and meta-analysis, found that mature-oocyte survival was 82.3% (602/731) following vitrification/warming and significantly better than slow cooling. Today, oocyte survival after vitrification is typically in the low-to-mid 90s%, though it varies across labs. Although effective, the approach may carry a cost: a large Australia/New Zealand registry study (2009–2023) found frozen-thawed oocytes had lower fertilisation rates and produced fewer usable blastocysts compared with fresh oocytes. Once an embryo does make it to transfer, though, the outcome gap closes for pregnancy and live birth rates, confirming clinical utility. The method is sound. What remains unsettled is how much of the shortfall before that point belongs to the biology and how much to the way the procedure is performed.

The Problem : A Protocol Built Around What Hands Can Do

Vitrification is a manual procedure run against a clock, though fast in its cooling step, it is a labor-intensive multi-step process. Standard protocols hold the cell in an equilibration solution for eight to fifteen minutes, then in the concentrated vitrification solution for about a minute, then plunge. Every transfer is performed by hand under a microscope. The equilibration window is itself a compromise, long enough for the cell to dehydrate and short enough to limit the toxicity of the solution doing the dehydrating, and the operator holds that trade-off in real time. The cooling rate, meanwhile, is set entirely by how quickly the operator can complete a single manual movement.

That is what the field is now trying to design around. A 2025 narrative review of shorter vitrification and warming protocols sets out two directions: abbreviated protocols and ultra-fast protocols.

Abbreviated protocols keep the same chemistry and shorten or reduce the steps. A 2025 study found comparable survival and pregnancy rates, with a lower miscarriage rate, using seven to eight minutes of equilibration rather than nine to ten. On the warming side the trend is one-step rehydration, with recent work documenting 1,266 healthy infants born this way at an overall live-birth rate of 51.3%.

Ultra-fast protocols push cooling rates higher instead, so that heat escapes almost instantly. Work published in 2019 showed that human eggs and zygotes are ready for ultra-fast vitrification after two minutes of cryoprotectant exposure, because the cell finishes dehydrating within about sixty seconds; longer exposure adds toxicity without adding benefit. In a germinal-vesicle oocyte model, ultra-fast vitrification with rapid elution reached 98.5% survival, statistically indistinguishable from fresh unfrozen cells, against 76.5% for conventional vitrification in the same model.

Both directions promise the same practical gains:

• less hands-on technician time

• less toxic chemical exposure

• more consistent results between operators

• an easier path to automation, and lower costs as clinics scale

Neither has become routine. Shorter windows and faster plunges demand a level of timing precision and repeatability, sustained across a full working day, that manual execution cannot reliably deliver.

Conceivable's Approach : Shorter and Faster Protocols with C:VIT

C:VIT, the vitrification workstation within AURA, our automation-assisted IVF laboratory, handles the two most delicate steps in freezing eggs or embryos: preparing them with cryoprotectant, and plunging them into liquid nitrogen.

A robotic pipetting system moves the sample step by step through cryoprotectant solutions, gradually raising the concentration so the cell dehydrates properly without being shocked. This part of the process is normally done by hand, under a microscope, with careful timing. C:VIT performs it using imaging to track the sample and computer-controlled precision to move it between solutions at exactly the right depth and timing. Built to follow the latest work on abbreviated protocols, its robotic construction also means the duration and number of cryoprotectant exposure steps can be modified as the evidence develops.

Once the sample is ready, a robotic arm plunges it into liquid nitrogen. This is where the ultra-fast part comes in: the system drives the sample down at high speed, using a proprietary plunging mechanism that carries it to the deep-frozen glass state in a fraction of a second.

In our own benchtop testing, we measured C:VIT’s cooling rate by taking temperature readings thousands of times per second using a tiny sensor, a thermocouple. That let us watch in fine detail exactly how fast the system could cool a sample. Across six tests, starting at room temperature of roughly 16°C, the sensor reached –196°C in about 5 milliseconds. A single blink of an eye takes about 100. The average cooling rate was therefore about 2.3 million °C per minute, with very little variation between runs, far exceeding published rates for standard manual devices, although we did not perform a direct comparison.

The Implication

Across six benchtop runs, the plunge produced the same cooling rate with very little variation. That reproducibility is the practical difference between a protocol that can be studied and one that can be run.

The literature on shorter protocols has established the direction: dehydration completes in about sixty seconds, longer exposure adds toxicity without benefit, and ultra-fast approaches achieve high survival on minimal cryoprotectant loads. Adopting any of them clinically requires the cooling step to be delivered the same way each time. C:VIT’s exposure steps are defined in software and can be reconfigured as those protocols are validated.

For clinic directors, what this offers now is a vitrification step with a measured, repeatable cooling rate. What it offers later depends on evidence that we can connect: whether shorter, less toxic protocols improve survival and development in patients.

Answer to the quiz

What cooling rate can automation-assisted vitrification achieve, and how does that compare to a traditional approach?

A. –3°C/min (the same as legacy slow freezing devices)

B. –12,500°C/min (enough to exceed the technical definition of “rapid cooling”)

C. –69,000°C/min (the same as manual vitrification devices)

✅ D. –2,300,000°C/min (an estimated 30x faster than manual approaches)

The answer is D, as we just discussed. Speed is the entire point of vitrification. The faster a cell cools, the less time damaging ice crystals have to form inside it. Our experiment is a direct, physical demonstration that C:VIT’s plunging mechanism achieves cooling fast enough to competitively support vitrification, and does so repeatedly.

The experiment measured the physical cooling capability of the machine using a bare temperature sensor, it did not test biological outcomes. It tells us the plunging system can deliver an extremely fast, consistent cooling rate; it does not, by itself, tell us how well real eggs or embryos survive and develop afterward. That’s a separate, downstream question that we now know it’s well worth exploring.

At the start of every IVF cycle, an embryologist picks up a dish and writes a name on it — a few characters in pen, scratched with a diamond tool, or fixed with an adhesive label. That mark, and the human chain of eyes that checks it at each step, is the foundation on which every subsequent safety check in the cycle is built.

A single sample mix-up in an IVF laboratory is not just a statistic. It is a family irrevocably altered legally, emotionally, and biologically. The problem is that the labelling step, as currently designed, is the one point in the workflow that no independent check can reach. Automation removes that potential point of failure by making the label and the dish the same thing, created together and inseparable from the moment of preparation onward.

The Problem : Identity Applied by Hand

Patient misidentification in IVF is rare. A 12-year retrospective analysis of over 36,000 treatment cycles and nearly 182,000 individual laboratory procedures at a large US IVF centre recorded recorded no major-grade non-conformances, the severity category that includes a confirmed pregnancy or birth following misidentification of gametes or embryos. The field’s safety record, relative to the complexity of what is being done, is genuinely strong.

That record reflects decades of increasingly careful process design around a labelling method that is largely manual. Every safety check that follows the labelling step assumes it was done correctly, though the labelling step itself is never independently verified. The mark itself can fail independently of the process: adhesive labels detach, handwritten ink smudges, diamond pen engravings produce illegible marks. The response has been to build verification around that step, starting with manual double-witnessing. The failure modes are well documented: checks omitted under time pressure, performed incompletely when staff are fatigued, or done out of sequence when workload peaks. In many laboratories, a single embryologist works alone for part or all of a session, making true independent witnessing structurally impossible.

Electronic witnessing systems go further. Each dish and tube is fitted with a radiofrequency tag carrying the patient’s identity, and readers positioned in every work area detect what is present at each bench. If material belonging to two different cases is in the same area at the same time, the system alerts the embryologist before anything can be handled. Over ten years of continuous use across 109,655 cycles and 849,650 individual witness points, alerts were triggered at a rate of 0.25% per witness point, holding steady across the decade. Of those, 144 were critical: cases where the material at risk of mixing was gametes or embryos. The word “triggered” here means they were caught and corrected, highlighting a need for the verification system. A separate seven-year dataset of 73,719 witness points found that critical alerts clustered in the late afternoon and around oocyte denudation, tracking working hours rather than technique.

Three limitations sit underneath those numbers. The tag identifies the vessel, not what is inside it, so when several embryos from the same patient are handled individually, RFID cannot tell one from another and manual double-witnessing remains necessary. Because tags can malfunction, dishes and tubes still carry a handwritten label as a fallback. And the system introduces a risk of its own: the authors note that operators can fall into witnessing samples without really looking, trusting the alert to catch what their attention no longer does. The most advanced identification system in routine use in IVF still rests on a handwritten label and a second pair of eyes.

There is another consideration apart from identity, about what the label is made of. IVF laboratories work hard to control the air their embryos develop in. Volatile organic compounds present in the room pass from the air into the oil overlay, then into the culture medium, and reach the embryo itself, where they can impair development. The Cairo Consensus, the field’s reference standard on laboratory air quality, lists solvents in labelling pens among the potential sources of such compounds, alongside cleaning agents, furnishings, and plasticware. Any material brought into the culture environment carries some version of this consideration. Laser engraving introduces none, because it adds nothing to the dish at all.

What is at stake in a cycle is well understood by the people going through it, and they say so when asked. In a study of over 400 IVF patients, 90% reported significant concern about the possibility of a sample mix-up, and 97% said they felt highly comfortable in a centre using electronic traceability for gametes and embryos. The demand for visible, verifiable traceability is already there.

Conceivable's Approach : Permanent Laser Identification and Automated Chain of Custody

C:DISH, our automated culture dish preparation system, incorporates a UV laser marking subsystem that engraves patient identification directly into the plastic of each dish before any media is added or any gamete contact occurs. The laser operates with a marking resolution of 1µm. Each dish receives two types of identifier etched onto its bottom and rim: a dot-matrix barcode readable by machine, and human-readable printed text. The barcode encodes the patient ID and, where multiple dishes are prepared for a single patient, a dish-specific ID that distinguishes each vessel in the set. Both are created in the same engraving pass.

Critically, the system reads these labels automatically. No manual scanning is required at any point. The operator can verify identity on screen at any time, but the default workflow requires no physical intervention: identity is confirmed by the system, not by a person reading a label and mentally confirming a match.

The engraving raises an obvious question: does marking the dish affect the surface the embryo will sit on? Optical coherence tomography, a technique that images the internal structure of a material through its full depth, gave a direct answer. The base of the dish is 1050µm thick, roughly the thickness of a credit card. At the validated power setting, the mark penetrates 800µm into it, leaving a quarter of a millimetre of untouched plastic between the engraving and the embryo-contact surface (Flores-Saiffe Farías et al., in review).

The structural question is one thing; the biological one matters just as much. The standard test is a mouse embryo assay, in which embryos are grown in the dishes under evaluation and the proportion reaching blastocyst is measured against a threshold the FDA sets at 80%. Laser-engraved dishes cleared that threshold comfortably, supporting blastocyst formation at rates between 90.5% and 100% against 90.5% in manually labelled controls (Flores-Saiffe Farías et al., in review). Embryos grown in engraved dishes developed no differently from those grown in dishes labelled by hand.

The closed-loop architecture extends beyond labelling. C:DISH is one station within the AURA workcell, and once a dish has been engraved and prepared there, it needs to travel: to C:SPERM for sperm preparation, to C:ICSI for fertilisation, or to C:CULTURE to rest, and so on through the entire workflow. Those movements are handled by C:HANDLER, a separate automated rig running behind the stations that picks up each dish, carries it to wherever the embryologist has directed it, and records the transfer on video. Throughout, the embryologist can see into each station and follow every dish on screen in real time, without ever handling it themselves. At each stop along the way, the system reads the dish and logs the step automatically. None of that would be possible without the mark made at preparation: it is what allows the custody chain to run continuously, with no person needed to confirm which dish is which at any point in the workflow.

The Implication

For the first time, the labelling step becomes something a laboratory can audit rather than assume. Every dish in every cycle carries a record of where it went, when, and under whose direction, and that record accumulates without anyone being asked to produce it, which matters because documentation that depends on staff finding time to complete it is documentation that degrades under pressure.

Regulation is moving in the same direction, if slowly. American Society of Reproductive Medicine (ASRM) guidance requires identification to be verified by two qualified witnesses, noting that an electronic system may serve that purpose depending on staffing and workflow. The expectation is clear even where the mechanism is left open, and laboratories that build the capability now will be ahead of the standard rather than retrofitting to meet it.

For clinic directors, the argument is straightforward. C:DISH closes the one step in the IVF workflow that no witnessing system could previously reach, and it does so without adding a task to anyone’s day. For patients, it is the reason they can trust that the embryo transferred is theirs.

Answer to the quiz:

In the IVF lab, which identity travels with the culture dish and can never be swapped?

A. A label stuck to the dish

OR

✅ B. A mark laser-cut into the dish

A label is a separate object attached to a dish by a person, whether adhesive, handwritten, or a radiofrequency tag. Anything attached can come loose, be applied to the wrong vessel, or be moved to another. A mark cut into the plastic cannot, because the identity and the dish are the same object.

That is what allows verification to happen automatically at every step: a machine can read the dish it has just picked up without a person confirming the match. Current practice is not unsafe, and misidentification in IVF is rare. What changes is where verification begins; at the moment identity is created, rather than at every point afterwards.

Leave a comment

An embryo spends its first five days sealed inside an incubator held at 37°C, in air with the CO₂ and oxygen fixed to a narrow band. To be looked at, it is lifted out into the open laboratory, and the moment the door opens the environment it was living in begins to slip.

The incubator’s only job is to hold that environment steady, because the embryo cannot hold it for itself. Every time an embryologist checks how the embryo is developing, the dish has to leave the incubator, and cooler, drier room air reaches it: the temperature falls, CO₂ comes out of the medium, and the pH drifts upward. The chamber then spends minutes rebuilding what it lost, and across five days this happens again and again, so the disturbance becomes a feature of the workflow rather than a lapse in it. The field’s cleverest answer has been the compartment and adding a camera to avoid drifts in culture conditions when assessing embryo development. A modern benchtop incubator hides four or more sealed chambers behind separate lids, so opening one patient’s dish never spills cold air onto another’s. A clinic may have to use several of these systems at once, while larger incubators are still commonly used for less sensitive work (like ensuring culture media are warm and ready to use). This footprint can become a bottleneck to scalability. Incubator automation changes two things: how many patient dishes one unit can hold , and how rarely a human hand has to reach inside.

The Problem

The environment inside an incubator is a substitute for the body. A preimplantation embryo develops with almost no ability to defend itself against a change in its surroundings, so temperature, gas and the pH of the culture medium are held for it: temperature at roughly 37°C, CO₂ at a set concentration because it governs the pH of the bicarbonate-buffered medium, humidity high enough to stop the droplet evaporating. The Cairo Consensus on IVF culture treats each of these as a determinant of the culture environment in its own right, to be maintained and monitored rather than assumed.¹

Opening the incubator interrupts all of them at once. How long that recovery takes depends heavily on the incubator, from a few minutes in a modern benchtop unit to considerably longer in a large-box incubator,² but the direction is always the same, and in a conventional laboratory it repeats at every assessment across the culture period.

Minimising embryo exposure to the environment remains a staple practice. When mouse embryos were cultured undisturbed for five days and compared against embryos exposed to two short (ten-second) door openings a day, the blastocyst formation rate did not differ between the groups (p=0.820), however disturbed embryos showed altered morphokinetics and formed blastocysts that were on average 1.562 μm smaller (p=0.016).³ In a non-randomised human sibling study of 285 IVF/ICSI cycles, reducing out-of-incubator embryo assessments from six to four increased blastocyst formation rates and quality, although clinical pregnancy and implantation rates were unchanged.⁴ Indeed, more recent work confirmed that big-box incubators with segmented doors, opened carefully, do not produce worse pregnancy rates.⁵

The Vienna Consensus (a set of recommendations drafted by IVF experts) sets blastocyst development rate as a key performance indicator, with a competency threshold of ≥40% and a benchmark of ≥60%, and names culture conditions as a determinant of that rate.⁶ But a development rate is an endpoint, and an endpoint can sit comfortably within benchmark while the path to it varies. The argument for stability is therefore a narrow one: a recognised determinant of embryo development could be allowed to fluctuate, and that fluctuation is real, measurable, and avoidable.

Incubator design has evolved largely to protect that environment, and two designs are in common use. Large-box incubators are the original: a single large chamber holding many dishes, but with one door, so each opening disturbs everything inside. Benchtop incubators answered that by splitting the space into six or more small sealed chambers behind separate lids, opening one patient’s chamber never spills cold air onto another’s; this is the current quality standard, though each unit holds only a few patients’ worth of embryos. Many benchtop units add a further refinement, time-lapse imaging, which photographs each embryo continuously inside its sealed chamber so it never has to come out to be looked at, removing the most frequent disturbance of all.

The pattern across all is the same trade-off: the better an incubator protects each embryo, the fewer patients it holds. A large-box unit takes roughly 10–15 patients but disturbs them together; a benchtop or time-lapse unit protects each one but typically holds only 2–6, so a busy clinic runs several at once, with large-box units kept for less sensitive work such as warming media. None of them removes the rest of the cycle. The embryo still has to leave the incubator for the steps that cannot be done inside it, insemination, handling, loading for vitrification, each other passage through open room air. Essentially, labs are being asked to balance capacity against environmental stability. A laboratory that runs on automation, processing many patients in parallel with minimal human handling, needs that same validated stability delivered at far greater capacity, and in a form a machine can access without a person opening a door.

Conceivable's Approach: A High-Capacity Automated Incubator, Validated Under Stress

C:CULTURE is the culture and incubation station within AURA, Conceivable’s automation-assisted IVF laboratory. It is a high-capacity automated incubator: where a standard IVF incubator holds embryos for roughly 10 to 15 patients, C:CULTURE holds around 50, with each patient’s dishes tracked individually as they move through the workflow. The system moves dishes in and out through a small automated port, so the chamber is reached by machine rather than by a person opening a large door.

Holding embryos for five times as many patients is only worth anything if each one sits in conditions as stable as a standard incubator provides. So the system was validated on three questions: does it hold steady in normal use, does it recover when something goes wrong, and do the embryos themselves develop normally inside it.

The first is the easiest to state. An embryo needs to sit at body temperature, 37°C, and stay there; drift of even half a degree is the kind of thing culture conditions are judged on. Across all five of the incubator’s racks, C:CULTURE held 37.0°C with a variation of three-hundredths of a degree. Then the harder test: opening it. A backdoor used to insert or remove dishes in the incubator was opened and closed 105 times in a row over 37 minutes, standing in for the repeated access of a busy day, and the temperature never drifted more than a few tenths of a degree from where it should be.

The second question is what happens in the situations a lab dreads. Left with its main front door wide open for three full minutes, the chamber cooled by two to three degrees, then climbed back to 37°C within the hour. Cut off from power entirely for an hour, it lost about 1.7°C and recovered over roughly two hours. Neither is a normal event; both show the environment bends under stress and returns, rather than breaking.

Temperature is only one of the things an embryo feels. The acidity of the fluid it grows in, its pH, has to stay in a narrow physiological band, because pH is set by the dissolved CO₂ the incubator controls; through both the open-door and power-loss tests, pH stayed inside that band. The concentration of the medium, its osmolality, has to stay in range too, since a droplet that slowly concentrates as water evaporates pulls the embryo out of balance; it held safe across a full seven days of culture. And the air itself has to be clean of the volatile compounds that off-gas from plastics and equipment; C:CULTURE kept these at the same low levels as a standard incubator and as the room air around it, and for microbes it tested to surgical-grade cleanroom standards.

None of this is worth anything if embryos do not actually grow. Across 275 mouse embryos, 96% reached the blastocyst stage by day 5 and 98% by day 6, with no meaningful difference from embryos grown in a standard incubator, and comfortably above the 80% mark used to certify a system as safe for embryos. Embryos grew just as well as in a standard incubator, at five times the capacity and inside an automated workflow.

The Implication

For a laboratory moving toward automated, high-volume work, the value is capacity that does not cost reliability. One validated system holds many patients’ embryos at once, roughly [50 against 10–15], which lets the rest of the automated workflow run at a scale a bench of separate incubators would struggle to match, and consolidates culture into a single monitored unit.

It also captures the environment continuously, in a structured form. Environmental data has long been treated as a compliance formality — sensors that raise an alarm when a threshold is crossed, logs that are filed and rarely read. Independent work is beginning to show how much that leaves on the table. Modelling continuous sensor data from two clinics, one recent study found that engineered environmental features (among them thermal stability and how quickly conditions recover after a disturbance) predicted pregnancy rates far better than simple averages, with temperature and CO₂ the most consistent signals across sites.7 The article points to where the field is heading: laboratories that can eventually ask which conditions matter, and by how much. The same move from retrospective checks to continuous monitoring is appearing on the process side, where near-real-time models have flagged laboratory performance shifts that fixed benchmarks missed.8

The incubator has long set a quiet ceiling on how many patients a lab can carry at once. C:CULTURE raises that ceiling from 10–15 patients to around 50, at a stability its validation put through repeated door openings and a simulated power loss. A laboratory that captures every condition its embryos experience, continuously and in the same structured form, is positioned to answer the question the field has never had the data to settle: which fluctuations actually change outcomes, and by how much. Every cycle it runs adds to that answer

Answer to the quiz:

How many patients can a single automated IVF incubator hold embryos for at once?

A. 10 to 15

B. 25 to 35

✅C. 45 to 55 or more

A conventional benchtop incubator, the current quality standard, holds only a few patients’ worth of embryos at a time (option A territory), which is why a busy lab runs several side by side. C:CULTURE holds embryos for around 75 patients [confirm] in a single automated unit, at a culture environment validated to the same standard, and reached by the system rather than by a hand opening a door. The point isn’t a warmer or steadier box. It’s the same protected environment, held for far more patients at once.

References

1. Cairo Consensus Group, 2020. ‘There is only one thing that is truly important in an IVF laboratory: everything’Cairo Consensus Guidelines on IVF Culture Conditions. Reproductive BioMedicine Online, 40(1), pp.33-60.

2. Swain, J.E., 2020. A Comparison of Embryo Culture Incubators for the IVF Laboratory. In Textbook of Assisted Reproduction (pp. 693-706). Singapore: Springer Singapore.

3. Nguyen, Q., Sommer, S., Greene, B., Wrenzycki, C., Wagner, U. and Ziller, V., 2018. Effects of opening the incubator on morphokinetics in mouse embryos. European Journal of Obstetrics & Gynecology and Reproductive Biology, 229, pp.64-69.

4. Zhang, J.Q., Li, X.L., Peng, Y., Guo, X., Heng, B.C. and Tong, G.Q., 2010. Reduction in exposure of human embryos outside the incubator enhances embryo quality and blastulation rate. Reproductive biomedicine online, 20(4), pp.510-515.

5. Kieslinger, D.C., Scholtenhave, S.C.M., Kostelijk, E.H., de Witte, S.F.H., Lambalk, C.B., Dreyer, K., van Eekelen, R. and Vergouw, C.G., 2026. Influence of incubator type on embryo development and clinical outcomes: A retrospective cohort study comparing time-lapse benchtop incubators versus big box incubators. European Journal of Obstetrics & Gynecology and Reproductive Biology, p.115160.

6. ESHRE SIG Embryology & Alpha Scientists in Reproductive Medicine. The Vienna Consensus. Reprod Biomed Online. 2017;35(5):494–510.

7. Asghari Varzaneh Z, Khoshkangini R, Saldeen P, Johansson L, Ebner T. Context-aware hierarchical Bayesian modeling of IVF laboratory environmental conditions. PMLR (ICML). 2026. arXiv:2606.20459.

8. Güell Penas E, Vazquez Vazquez I, Ochando Sanchez I, Vazquez Vazquez J, Fernandez Tarjuelo M. Near–real-time adaptive ML for continuous quality monitoring in IVF laboratories. Hum Reprod. 2026; abstract deag083.261.

Intracytoplasmic sperm injection, or ICSI (pronounced “ick-see”), is one of the most technically demanding procedures in reproductive medicine. An embryologist sits at a specialised microscope, uses glass needles finer than a human hair to select a single sperm, immobilise it, and inject it directly into a mature egg. The target is a cell roughly 0.1 millimetres across, invisible to the naked eye. Success depends on precise control of angle, force, and depth, adjusted in real time by the practitioner’s judgement.

Today, most IVF cycles worldwide involve ICSI. And for over thirty years, essentially every injection has been performed by a human hand. That is starting to change.

Why ICSI Matters, and Why It's So Difficult to Standardize

ICSI was developed to address cases where sperm cannot fertilize an egg on its own, including severe male infertility, previous failed fertilization attempts, or situations where only surgically retrieved sperm is available. Rather than hoping sperm will find and penetrate an egg in a dish, ICSI delivers fertilization directly: one sperm, one egg, one controlled injection.

The results transformed IVF outcomes. Fertilization rates that were previously unpredictable became much more consistent. Since its first successful use in 1992, it has become the dominant fertilization method in IVF. Today, ICSI accounts for the majority of IVF treatments carried out in developed countries. It is estimated the approach is used in 80% of IVF treatments in the US.

But ICSI carries a catch that is rarely discussed outside laboratory walls. The procedure is almost entirely dependent on the skill and consistency of the individual embryologist performing it. The Vienna Consensus, the authoritative performance standard for IVF laboratories, sets the acceptable rate of oocyte damage during ICSI at no more than 10%, with best practice defined as 5% or below ¹. Those figures reflect the fact that the procedure does carry real risk of harm to the egg, and that risk is directly tied to how the injection is performed.

Two embryologists performing ICSI at the same clinic, using the same equipment and the same media, can produce measurably different outcomes ². The same embryologist can produce variable results depending on the position of a particular egg, or the quality of a particular batch of needles, or equipment calibration. These are not criticisms of individual practitioners. They are inherent features of any manual procedure where the instrument is a glass needle and the target is to capture the smallest human cell (the sperm cell).

The Case for Piezo-ICSI

Before we get to automation, it is worth understanding a variation on the standard technique that is already reshaping how injections are performed.

In conventional ICSI, the embryologist uses suction from within the needle to rupture the egg’s inner membrane (the oolema) before depositing the sperm. This momentary aspiration of cytoplasm, the fluid inside the cell, is what confirms the membrane has been breached. It works reliably, but the technique demands significant skill and places mechanical stress on the egg.

Piezo-ICSI uses a different approach. A tiny electrical pulse sends a controlled vibration along the injection needle, allowing it to pass through the egg’s membranes without the need for cytoplasmic aspiration. The needle penetrates, the sperm is deposited, the needle withdraws. The egg experiences much less mechanical disturbance.

The evidence for piezo-ICSI is accumulating. A prospective study by Zander-Fox and colleagues (2021) ³, the first conducted outside Japan and at a centre with no prior piezo experience, found that piezo-ICSI produced a fertilization rate of 80.5% compared with 65.8% for conventional ICSI (adjusted model, p < 0.0001). Oocyte degeneration fell from 8.6% to 4.4%. Perhaps most importantly for patients, the number of embryos available for transfer or freezing increased by roughly one per cycle on average.

A previous study by Furuhashi and colleagues (2019) ⁴ used a sibling-oocyte design, dividing eggs from the same patients between piezo and conventional ICSI, which elegantly removes patient-level variables. In women over 35, piezo-ICSI produced a fertilization rate of 74.8% versus 64.3% for conventional injection, and significantly higher blastocyst formation at day five or six (52.4% versus 39.6%). In women under 35, results were equivalent, suggesting piezo’s advantage is most pronounced when egg quality is already under pressure from age.

The reason piezo-ICSI has had limited uptake for years is partly the use of mercury inside the needle as a damping fluid, which created occupational health and embryotoxicity concerns. Modern protocols now use medical-grade alternatives including perfluoro-n-octane, a compound already approved for ophthalmic use, which resolves that barrier. Damping-fluid-free approaches are also being developed.

Where Automation Enters

Piezo-ICSI has one feature that makes it particularly well suited to automation: the moment of membrane penetration is triggered by a programmable electrical pulse rather than by tactile judgement. That single change transforms the most difficult sub-task in ICSI, the one hardest to teach and hardest to replicate mechanically, into something that a computer can initiate with precise, reproducible parameters.

This insight is at the heart of what Conceivable Life Sciences has built with the C:ICSI module, part of AURA, its automation-assisted IVF laboratory.

A landmark proof-of-concept study by Mendizabal-Ruiz and colleagues ⁵, published in Reproductive BioMedicine Online and awarded the Robert G. Edwards Prize Paper Award 2025 by RBMO in recognition of its significance, demonstrated that a fully automated, remotely operated system could execute all 23 micromanipulation steps of ICSI. The system achieved 80% fertilization from the five eggs it injected, comparable with 100% from the three manual controls. A frozen embryo transfer from the automated group resulted in a healthy baby boy.

The operator who guided most of the remote steps was located in Hudson, New York. The eggs were in Guadalajara, Mexico. The distance between them was approximately 3,700 kilometres.

What AURA's Automation-assisted ICSI Actually Does

Understanding why this matters requires a closer look at what the AURA system does at each stage of the injection, because the innovation is not simply that a machine holds the needle.

• Sperm selection and identification. The system uses a computer vision tool to identify sperm cells in real time [as we discuss here], ranking them based on movement patterns associated with developmental potential. The selected sperm is tracked as it moves, and the system centres it in the microscope’s field of view automatically.

• Laser immobilization. Rather than manually immobilizing the sperm by bending its tail with the needle, the system fires a calibrated laser pulse at the midpoint of the sperm’s tail. This is precise, repeatable, and leaves the sperm head undamaged (Ebner 2002) ⁶.

• Oocyte positioning and zona penetration. Computer vision identifies the egg at multiple magnifications and aligns the tools (the holding pipette and injection needle). The piezo pulse, with controlled amplitude and frequency, then drives the needle through the zona pellucida (the egg’s outer shell) and the oolema with defined depth parameters.

• Verification of sperm deposition and oocyte integrity. This is perhaps the most significant quality assurance advance. In manual ICSI, successful injection is confirmed by the embryologist’s visual impression during the procedure. AURA’s system uses computer vision to verify that the sperm has been deposited inside the egg, and separately confirms that the egg has survived the injection intact. This creates objective, documentable evidence for each injection.

A study published in Human Reproduction by Chavez-Badiola and colleagues, placed ICSI within a full day automation sequence ⁷. Across 11 patients treated, the automated pathway achieved 64.3% fertilization and 42.2% usable blastocyst formation. Five healthy babies were born from nine patients with positive pregnancy tests. The Chief Editor of Fertility & Sterility has described this as a turning point in the history of IVF lab automation (Paulson 2026) ⁸.

Piezo-ICSI performed with automation-assistance. A software interface controls injection speed and recognises the egg, the sperm cell, and the micromanipulation tools. Individual steps are monitored and recorded (image credit Mendizabal-Ruiz 2025)

What Automation Cannot Yet Replace

It is important to be clear about the current state of the technology. In the Mendizabal-Ruiz study, the automated system completed 49.6% of the required micromanipulation steps independently. The remainder were completed under digital control by the remote operator. Each injection took an average of nearly ten minutes, considerably longer than manual ICSI in routine clinical use.

In the full day automation sequence study, operators needed to intervene during egg handling and ICSI approximately a third of the time. The vision system occasionally misidentified a structure; an egg might require repositioning; cumulus cells sometimes required manual removal despite enzymatic treatment.

These are the honest limitations of a technology still in its early clinical phase. They also represent a clear roadmap for where engineering effort is now focused, including higher autonomy rates, faster cycle times, and expanded training datasets across patient populations and clinical settings.

The current system removes the subjective element from the moment of membrane penetration. The piezo pulse fires at defined parameters. The depth is controlled. The verification is machine-generated. Those properties do not depend on experience, shift length, or the particular difficulty of one patient’s eggs.

Why Standardization Matters Beyond Individual Clinics

There are around 2,900 IVF clinics operating in the United States alone, and many more globally. Their outcomes vary considerably, shaped by local embryologist experience, equipment, culture media, and protocols. A patient undergoing IVF at a top-tier academic centre has access to capabilities that a patient in a rural area or a lower-income country may not.

Automation addresses this asymmetry in two ways. First, a standardized automated system performs each step within defined parameters regardless of where the equipment is installed, which removes the ceiling on how consistently a less-experienced clinic can perform. Second, as the remote ICSI demonstration showed, expert oversight can be extended across distance, allowing a specialist to supervise or intervene from anywhere with a stable internet connection.

The clinical relationship between patients and their care team remains central. What changes is this: the quality of a specific technical procedure, the injection of a single sperm into a single egg, no longer depends on whether the best embryologist in the country happens to be at work that morning.

Answer to the Poll

Which ICSI method provides more consistent outcomes with verified completion?

A. Manual ICSI : hand controlled,
(relying on practitioner skill for injection force, angle, and real-time visual confirmation)

OR

✅B. Automated-assisted ICSI,
(with piezo-assisted membrane penetration, controlled depth, and computer vision verification)

Based on the evidence discussed here, the answer is B, with an important nuance. First-generation automation-assisted prototypes have not yet matched skilled embryologists on fertilization and blastocyst rates. . What they have demonstrated, consistently and documentably, is standardized execution of the injection itself: a controlled piezo pulse, defined depth parameters instead of visual estimation, and automated confirmation of sperm deposition.

Five babies born from a proof-of-concept study, and one from the remote ICSI study, are proof that automated systems can safely participate in clinical IVF pathways. The next chapter, underway now through the expanded AURA clinical trials, is establishing how consistently and at what scale that participation can occur.

References

1. E.S.I.G. and Alpha Scientists in Reproductive Medicine, 2017. The Vienna consensus: report of an expert meeting on the development of ART laboratory performance indicators. Reproductive BioMedicine Online, 35(5), pp.494-510.

2. Tiegs, A.W. and Scott, R.T., 2020. Evaluation of fertilization, usable blastocyst development and sustained implantation rates according to intracytoplasmic sperm injection operator experience. Reproductive BioMedicine Online, 41(1), pp.19-27.

3. Zander-Fox, D., Lam, K., Pacella-Ince, L., Tully, C., Hamilton, H., Hiraoka, K., McPherson, N.O. and Tremellen, K., 2021. PIEZO-ICSI increases fertilization rates compared with standard ICSI: a prospective cohort study. Reproductive biomedicine online, 43(3), pp.404-412.

4. Furuhashi, K., Saeki, Y., Enatsu, N., Iwasaki, T., Ito, K., Mizusawa, Y., Matsumoto, Y., Kokeguchi, S. and Shiotani, M., 2019. Piezo‐assisted ICSI improves fertilization and blastocyst development rates compared with conventional ICSI in women aged more than 35 years. Reproductive Medicine and Biology, 18(4), pp.357-361.

5. Mendizabal-Ruiz, G., Chavez-Badiola, A., Hernández-Morales, E., Valencia-Murillo, R., Ocegueda-Hernández, V., Costa-Borges, N., Mestres, E., Acacio, M., Matia-Algué, Q., Farías, A.F.S. and Carreon, D.S.M., 2025. A digitally controlled, remotely operated ICSI system: case report of the first live birth. Reproductive BioMedicine Online, 50(5), p.104943.

6. Ebner, T., Moser, M., Yaman, C., Sommergruber, M. and Tews, G., 2002. Successful birth after laser assisted immobilization of spermatozoa before intracytoplasmic injection. Fertility and sterility, 78(2), pp.417-418.

7. Chavez-Badiola, A., Mendizabal-Ruiz, G., Flores-Saiffe Farías, A., Costa-Borges, N., Murray, A., Alikani, M., Silvestri, G., Millan, C., Hernández-Morales, E., Valencia-Murillo, R. and Medina, V., 2026. Automated oocyte retrieval, denudation, sperm preparation, and ICSI in the IVF laboratory: a proof-of-concept study and report of the first live births. Human Reproduction, 41(2), pp.214-230.

8. Paulson, R.J., 2026. Automation in the gamete laboratory: have we reached a turning point?. F&S Reports, 7(3), pp.171-172.

During egg retrieval, a doctor draws the fluid out of each follicle in the ovary, the small sac where an egg has been developing, and that fluid is sent to the laboratory with the egg somewhere inside it. The egg is about the width of a human hair and it does not float free: it sits wrapped in a cloud of supporting cells, a structure called the cumulus-oocyte complex, which an embryologist locates under a microscope and moves into culture. Finding it can be difficult because the same aspiration sometimes draws up blood and often small fragments of tissue, and against that cloudy background a complex can be hard to tell apart from everything around it. Automated optical scanning changes what the search depends on: instead of what happens to be visible in a single section of the dish, it images the full depth of the fluid, so a complex can be captured in the image even when it is hard to distinguish by conventional microscopy alone.

The Problem: The Eggs That Are Never Found

The number of eggs collected in an IVF cycle is one of the strongest predictors of whether it ends in a birth. A UK analysis of more than 400,000 treatment cycles found live birth rising with the number of eggs retrieved ^[1]. Every complex found is therefore a chance at an embryo, and every complex missed is a chance removed. That arithmetic is hardest on patients who produce only a few eggs, where a single one can decide whether there is an embryo to transfer or none at all.

The miss is not hypothetical. When one group processed follicular fluid that embryologists had already searched and set aside, using a method that captures eggs physically rather than by eye, it recovered at least one additional egg in more than half of cases, across 61 embryologists at four clinics ^[2]. The recovery held in samples that had been screened twice, and in patients with five eggs or fewer it still found extra eggs about a third of the time. Some laboratories already re-search the discarded fluid of low-count patients for this reason. The eggs still slip through, because the limit is the fluid and the microscopy available, not the care taken with it.

Part of the difficulty is optical. A dish of follicular fluid has depth, and complexes may become obscured by what floats above them.

A prolonged search carries a second hidden cost. While a complex sits in follicular fluid on the bench, it is in an unbuffered medium open to the air, and its environment drifts. Aspiration alone cools the fluid from about 37°C toward 29°C ^[3], and an oocyte held below temperature can show a disrupted meiotic spindle that may not fully recover on rewarming ^[4]. Follicular-fluid temperature at retrieval, when it stays within a narrow band, tracks with later blastocyst development and live birth ^[5], and laboratory guidance calls for aspirate searching to hold appropriate conditions or be completed without delay ^[6]. The longer a complex waits to be found, the longer it spends in conditions the incubator exists to prevent.

And the miss leaves no record. An embryologist cannot account for an egg they never saw, so the loss is invisible, and the field has no measure of how many eggs are recovered against how many were present.

Conceivable’s Approach: Imaging the Whole Dish, Not Scanning the Surface

At Conceivable, our approach has been to replace the classic microscope with a more comprehensive imaging setup. C:EGG, the oocyte-detection workstation within AURA, the world’s first automation-assisted IVF laboratory, uses a wide field of view digital camera to observe the entire follicular fluid dish in one high-resolution image, supported by automation-assisted detection of the cumulus-oocyte complexes present. When needed, a second imaging system captures a stack of images at successive depths through the region of interest, piercing through the layers of fluid, cells and debris (building what is technically called an image stack). Because the whole volume is recorded in a single pass, an oocyte can be identified wherever it sits in the observed region, rather than only when it happens to fall on the plane in focus. Software marks every candidate complex on screen, and the embryologist confirms each one, triggering automated picking.

A high-resolution digital camera is used to capture an entire dish containing follicular fluid, with dedicated software pinpointing the location of probable oocytes (green squares).

The hardest complexes to read are the opaque ones, where dense cumulus or blood scatters and absorbs the light a microscope depends on triggering the use of the Z-stack approach we mentioned before. The technology used to achieve this is optical coherence tomography (OCT), the same depth-resolved imaging principle used routinely in eye clinics. In this application OCT uses near-infrared light (low energy light unlikely to damage the egg [7], which passes through cumulus and blood where visible light cannot, and builds a cross-sectional image from what reflects back, resolving the oocyte inside a complex that looks opaque under a standard microscope. Its purpose here is confirmation: establishing whether an oocyte is present in a well that the eye cannot read. Reading an oocyte’s structure to judge how it might develop is a separate question OCT may one day help answer, but it is not a claim we make here.

Left: a cumulus-oocyte complex in clouded follicular fluid under conventional brightfield microscopy, where the oocyte cannot be easily made out. Right: an optical coherence tomography cross-section of the same complex, where near-infrared light passes through the cumulus and debris to resolve the oocyte inside. OCT confirms that an oocyte is present.

In this example, OCT has been used with a lower magnification lens and wider field of view to capture an illustrative image of ten denuded eggs. The z-stacks were used to build a full 3D render of the dish surface with the eggs sitting on top.

The principle holds up where it matters most. In a feasibility study using bovine follicular fluid, across 149 complexes where the presence or absence of an oocyte inside was later confirmed by denudation, the OCT-based assessment correctly identified an oocyte in 97.8% of cases and never flagged an empty well, missing 3 oocytes where the best-performing expert assessor relying on classic microscopy images missed at least 13 and a beginner missed as many as 63. Almost half the samples were clouded by blood or dense cumulus, the conditions that defeat the eye, and the system held its accuracy through them. These are animal-model results, and validation in human samples is the necessary next step; what they establish is that the limit on finding oocytes in opaque fluid is the reach of visible light, and that imaging past it recovers what would otherwise be lost.

The Implication

Completeness stops being a property of the search and becomes a property of the instrument. Every dish receives the same full-width scan and every candidate the same characterisation, whether it is the first retrieval of the day or the last.

This does not reveal how many eggs a dish truly holds. What it changes is how many are found, and that the finding no longer varies with who is at the microscope: the same sample can be read differently by different people, whereas the system reads it the same way regardless of operator, workload, or hour of the day.

For the patient who produced three eggs, the question is no longer how carefully the fluid was searched, but whether every layer of it was imaged; and that is now a property of the instrument, not a demand on the person at the microscope.

The Answer to the Quiz:

Which tool is best suited to find every oocyte in a dish of follicular fluid?

A. A conventional transmission microscope

OR

✅B. A camera imaging the full depth of the dish in stacked layers, top to bottom

What decides whether an egg is found is whether it was visible at the moment the dish was examined, and sometimes classic microscopy cannot identify samples through blood and dense cellular debris. Imaging the full depth of candidate structures in near-infrared light removes that limit for almost every complex, though not yet quite all (yet). The point is not that detection becomes perfect, but is measurably improved and stops depending on what visible light happens to reach the eye.

References

1. Sunkara SK, Rittenberg V, Raine-Fenning N, et al. Association between the number of eggs and live birth in IVF treatment: an analysis of 400,135 treatment cycles. Hum Reprod. 2011;26(7):1768–1774.

2. Mutlu BR, Civale SC, Diettrich J, et al. Microfluidic automation improves oocyte recovery from follicular fluid of patients undergoing in vitro fertilization. Nat Med. 2026;32(3):906–914.

3. Redding GP, Bronlund JE, Hart AL., The effects of IVF aspiration on the temperature, dissolved oxygen levels, and pH of follicular fluid, J Assist Reprod Genet. 2006;23(1):37-40.

4. Wang WH, Meng L, Hackett RJ, et al. Limited recovery of meiotic spindles in living human oocytes after cooling-rewarming observed using polarized light microscopy. Hum Reprod. 2001;16(11):2374–2378.

5. Sherbahn R. Assessment of effect of follicular fluid temperature at egg retrieval on blastocyst development, implantation and live birth rates. Fertil Steril. 2010;94(4 Suppl):S68–S69.

6. Practice Committees of ASRM and CAP. Comprehensive guidance for human embryology, andrology, and endocrinology laboratories: a committee opinion. 2022.

7. Aquilina M et al. Investigating phototoxicity of optical coherence tomography imaging in porcine and human spermatozoa. Reproductive BioMedicine Online, 2025; 52

In an IVF laboratory, an embryologist sits at a microscope with a dish of prepared sperm and a task: scan the moving cells, choose one, and inject it directly into a waiting egg. The sperm they select carries half the future embryo’s genetic material, a decision the embryologist is trained to make, guided by established criteria for what predicts which cell is most likely to fertilise successfully. That choice belongs to the embryologist. What automation-assisted support adds is objective, quantified data on the parameters that are hardest to assess at the microscope, available at the moment the decision is made.

The Problem: Seeing and Measuring

For IntraCytoplasmic Sperm Injection (ICSI), the most widely used fertilization technique in IVF worldwide, a single spermatozoon is selected by the embryologist and injected directly into the egg. That selection happens at the microscope, with a prepared sperm sample that has been concentrated and processed to remove debris and non-motile cells [see: Gentle Sperm Preparation]. The embryologist scans the field and assesses each candidate sperm against established criteria, chiefly the shape of the head, the integrity of the midpiece and tail, and the character of the cell’s movement, whether it travels forward progressively, with sufficient speed and directionality to indicate fertilisation potential.

Those judgements are rooted in the same criteria used earlier in the process, when the semen sample is first examined in the andrology laboratory to assess male fertility. There, the embryologist counts sperm one by one and tallies how many meet the criteria for normal form. This figure carries direct clinical weight: the WHO sets the lower reference limit at 4% normal forms. In most laboratories, the count runs to 200 sperm.¹

But how accurate is this assessment? The problem is statistical: 200 sperm is a sample, and every sample carries uncertainty. Imagine a jar holding millions of beads, most blue, a small fraction red. You want to know what fraction is red, but you cannot count them all — so you scoop out 200 and count those. You find 8 red beads: 4%. Reach in again, and you might find 6, or 11, or 5. The jar has not changed, your handful has. This is what happens every time an embryologist counts 200 sperm. Each count is performed correctly. The result is still an estimate and because of the randomness in which 200 cells were examined, that estimate carries an uncertainty range of roughly ±2 to 3 percentage points either side of the true value. A reported result of 4% could reflect a true proportion anywhere between 2% and 8% in the same sample. This has been quantified precisely. The difference between 3% and 5% is clinically meaningful: it can be the line between a normal sample and an abnormal one. To reliably tell those apart, an observer would need to count closer to 1,500 sperm;² roughly eight times the routine standard, something no clinical laboratory does in practice.

That uncertainty compounds when considering operator variability. Picture two embryologists, each given the same sample to assess independently, by the same method. When they compare results, they will not land on the same number. This ‘inter-observer variability’ in sperm morphology assessment is a well-documented limitation in reproductive medicine. A study of 62 trained staff, each assessing the same sample independently, found that the estimates of the percentage of normal sperm ranged from 6% to 39% across participants.³ Experienced staff performed closer to the reference value, and following recommended methodology showed lower variability, but the variability persisted across all levels of expertise. It is a structural property of visual assessment, not a deficit in training: human observers applying agreed criteria to the same sample do not consistently produce the same result.

When it is time to select a single sperm for ICSI, the embryologist is no longer doing the slow, sperm-by-sperm assessment, but forming a judgement in seconds, under time pressure, about a moving target that is the smallest cell in the human body. But two embryologists scanning the same field need not settle on the same cell, the subjectivity of visual assessment carries from the bench to the moment of selection, and which cells happen to catch the eye shapes which one is chosen. What real-time visual assessment cannot reach is precise quantification: how fast a cell moves in a straight line, how linear its path is, the specific character of its head movement, and how it compares against its neighbours. These parameters: straight-line velocity (VSL), path linearity (LIN), and head movement pattern (HMP), are each independently associated with fertilization potential and blastocyst formation.⁴ They are also, by their nature, computational: determining them requires tracking a cell’s trajectory frame by frame and calculating values from that track. Scale compounds all of this. In standard practice, the full ICSI procedure takes approximately 90 seconds per oocyte of which sperm selection, immobilisation and pick-up account for around 40 seconds (as we previously reported)5. In laboratories performing a detailed morphological selection of each candidate sperm at high total magnification (1,000x vs routine 200-400x for ICSI), a single 10-oocyte case takes approximately 2.5 hours.6

Conceivable’s Approach : A Measurement Layer at the Point of Selection

At Conceivable, we have integrated SiD, Sperm ID, a motility quantification tool, within AURA, our automation-assisted IVF laboratory. SiD measures sperm motility in real time, at the ICSI dish, at the moment the embryologist is making their assessment, and provides objective, quantified data to inform that decision.

SiD watches the same field the embryologist is watching, through the microscope’s camera, and tracks each individual sperm cell at fifteen frames a second, reconstructing each one’s path and computing for each cell the three motility parameters we mentioned above: straight-line velocity (VSL), the linearity of its curvilinear path (LIN), and head movement pattern (HMP), finishing its assessment of all cells in under a second. From these three measurements, SiD generates a categorical score for each sperm: Low, Medium, Good, or Best. It displays a ranked recommendation on screen in real time of the top three sperm simultaneously. The embryologist reads it, uses that information alongside their own assessment, and selects the cell to inject.

The SiD interface during live ICSI. The field shows the ICSI dish as viewed under the microscope; each sperm under assessment is marked with a circle. Motility parameters are evaluated in real time. SiD assesses the entire field of view and highlights up to three best-performing spermatozoa simultaneously, and ranks them on the left panel from optimal to suboptimal with individual parameter scores displayed alongside each candidate.
Images: Courtesy of IVF 2.0 Ltd. Interface: SiD v1.0, now built into AURA automation-assisted IVF laboratory.

The SiD v1.0 measures movement, and only movement, with expanded capability including morphological classification (ie, assess the shape of a cell) having been recently deployed with SiD v2.0, although most of the available peer-reviewed data has been collected on SiD v1.0. So the case for SiD today is the motility case: a repeatable figure for how a cell moves which cannot be measured by eye. Additionally, SiD analyses sperm already present in the ICSI dish at the moment of selection. The assessment is non-invasive and integrates into the existing automation workflow without any modification to the sperm or the ICSI procedure itself.

The figure SiD returns stays consistent no matter who is at the microscope, how many dishes have come before, or how late in the day it is, reducing observer variability. It also helps assess a larger and more balanced mix of sperm cells: by looking at every cell visible in the field, its recommendation does not depend on which ones drew attention in the moment. If that consistency is real, it makes a testable prediction: SiD should never do worse than a skilled embryologist working alone, and where it helps, it should help most exactly where human variability or biological difficulty runs highest; with less experienced operators, and with lower-quality eggs that leave less room for a poor choice to be forgiven. The evidence for that now runs to four independent studies, carried out in Canada, Japan, Spain and Austria, across more than 1,600 mature oocytes across different laboratories, different sperm preparation methods, different patient populations.^7-10 If standardisation is real, the benefit should hold across settings rather than belong to one lab’s way of working.

Across all four, the pattern runs one way. SiD-assisted sperm selection before ICSI was at least as good as conventional selection everywhere it was tested, and where it pulled ahead it did so exactly where the argument said it would. In less experienced hands, the gap between junior and senior embryologists closed — with SiD, junior performance became statistically indistinguishable from senior.⁹ And in older patients, where eggs have less biological capacity to compensate for a poor sperm choice, blastocyst formation was markedly higher when SiD’s top-ranked sperm were used.⁷ No study has reported harm.

The Implication

The studies asked a narrow question: whether objective motility data changes outcomes at the point of selection. The larger question is what changes once the measurement exists at all.

Sperm selection has always been a private skill built over years at the microscope, held in the hands and eyes of individual embryologists, and becoming more consistent with experience.³ That is part of why the field can hold pockets of excellence. A skill that is difficult to transfer is difficult to improve on collectively; each lab starts again from its own people. A measurement behaves differently. A figure for how a cell moves means the same thing in Valencia as in Osaka, and a standard that holds across operators and sites is the precondition for the field learning as a field, for one lab’s result becoming a benchmark another can be held to, rather than an anecdote about a particularly good embryologist.

And every measured selection leaves a record that did not exist before: these motility values, this embryo, this outcome. One clinic’s worth is a logbook; many clinics measuring the same way is a dataset and the basis for learning which parameters carry weight, and by how much. For the first time, the act of selecting sperm will produce the evidence that can improve it. And that evidence will widen as the measurement does: motility yielded first because it is dynamic and computable, a path traced across frames; morphology is its natural complement, and a new version that quantifies it has already been developed.

A measurement that produces the same standard of selection regardless of who is holding the pipette means a rapid pathway to skill sharing. But the more immediate argument is for the patient in the room. The evidence points consistently toward one group in whom selection support counts most: older women, using their own eggs, where the biological margin for a poor sperm choice is already narrow. These are not the easiest patients. They are the ones for whom IVF is most likely to fail, and for whom each cycle carries the most weight. If objective motility data raises blastocyst formation in that group — and the current evidence suggests it does — then the case for using it is not that it replaces the embryologist’s judgement. It is that it gives that judgement something it has never had before: a measurement, at the moment it matters most.

The Answer to the Quiz:

Which approach to single-sperm selection gives more consistent information for ICSI?

A.) A trained embryologist’s eye

B.) Objective software-based data

✅ C.) Both, used together

The answer is both. The eye is reliable for some things and structurally blind to others. An embryologist can tell at a glance whether a sperm is moving progressively, whether it looks vigorous, whether its structure is intact. What no eye can do, at any level of experience, is put a precise figure on how fast a cell travels in a straight line, how linear its path is, or how it compares with the hundreds of others moving in the same field, in the few seconds before the egg is injected. So selection has largely run on the eye alone. Not because a second, objective input wouldn’t help, but because it hasn’t existed in real time, at the dish, in the moment the decision is made.

That is the gap SiD is built to close. It provides objective, quantified motility data on each cell in the field, available at the point of selection — set alongside the embryologist’s own judgement rather than in place of it. The decision stays with the embryologist. What changes is that they no longer have to make it on impression alone.

Our automation-assisted lab is built to do just that.

References

1. World Health Organization. WHO laboratory manual for the examination and processing of human semen. 6th ed. Geneva: WHO; 2021.

2. Björndahl L. What is normal semen quality? On the use and abuse of reference limits for the interpretation of semen analysis results. Hum Fertil. 2011;14(3):179–186.

3. Eustache F, Auger J. Inter-individual variability in the morphological assessment of human sperm: effect of the level of experience and the use of standard methods. Hum Reprod. 2003;18(5):1018–1022.

4. Mendizabal-Ruiz G, Chavez-Badiola A, et al. Computer software (SiD) assisted real-time single sperm selection associated with fertilization and blastocyst formation. Reprod Biomed Online. 2022;45(4):703–711.

5. Mendizabal-Ruiz G, et al. A digitally controlled, remotely operated ICSI system: case report of the first live birth. Reprod Biomed Online. 2025 May;50(5):104943.

6. Cherouveim P, Velmahos C, Bormann CL. Artificial intelligence for sperm selection — a systematic review. Fertil Steril. 2023;120(1):24–31.

7. Carrión-Sisternas L, et al. Automated AI for real-time sperm selection in ICSI: reducing variability and studying the role of sperm in embryo development. Reprod Biol Endocrinol. 2025;23:155. [Discussion section re: future morphology capability and registered trial NCT07163754]

8. Montjean D, et al. Automated Single-Sperm Selection Software (SiD) during ICSI: A Prospective Sibling Oocyte Evaluation. Med Sci. 2024;12(2):19.

9. Nakano S, Okabe M, Fujita M, Takahashi K. A comparative study of AI-based automated sperm selection and embryologists: evaluation of sibling oocyte outcomes in intracytoplasmic sperm injection. Hum Reprod. 2025;40(Supplement_1):deaf097.227.

10. Pastor Leary C, et al. The role of artificial intelligence (AI) in selecting sperm for intracytoplasmic sperm injection (ICSI): a pilot study. Hum Reprod. 2025;40(Supplement_1):deaf097.376.

Every IVF cycle generates thousands of biological signals: temperature shifts, timing intervals, environmental fluctuations, procedural decisions, each one potentially relevant to what happens to the embryo. Most of it is carried in an embryologist’s hands and memory, and when the cycle ends, most of it disappears.

The IVF lab generates far more biologically meaningful information than it captures. Manual workflows record what a person notices, at preset intervals: a morphology score, an environmental log, a procedural note entered at the end of a session. The result is a record of endpoints, not process. The fluctuations, the intervals, the decisions made between observations go largely undocumented. Automation assistance is built on a different premise: that capturing the full record of a cycle should be a function of how the lab runs, built into the workflow rather than added on top of it.

The Problem : A Capture Gap, Not A Data Gap

The IVF laboratory has, over the last two decades, accumulated impressive point solutions for capturing what happens to gametes and embryos. Time-lapse incubators generate continuous imaging and environmental records for the embryos cultured inside them. Electronic witnessing systems log identity-critical handling events. Quality management systems track media lots, incubator alarms, and equipment calibration. Individually, each addresses a real gap. Collectively, they do not yet constitute a process record. The IVF laboratory is an information-rich environment, it generates thousands of biological signals per cycle. It needs the infrastructure to capture, structure, and retain them. A typical cycle generates a small set of recorded observations such as fertilization rate, cleavage stage, blastocyst grade, transfer outcome. Between those checkpoints, the lab is effectively running blind. Data points such as temperature fluctuations during dish handling, timing intervals between procedural steps, and subtle differences in technique between operators are not captured systematically. As the field moves toward greater transparency and auditability, a complete process record covering everything that happened to a patient’s biological material, becomes an expectation rather than an exception. In a manual lab, that record is incomplete, and often inaccessible.

The consequences play out across three distinct moments. During a cycle, the clinical team cannot see what is actually happening between observations: the temperature that shifted during a handling step, the interval that ran longer than protocol, the technique that varied between operators. After a cycle, if outcomes are poor, there is no process record to investigate. The team cannot necessarily distinguish a biological failure from a process failure, or trace cause with credible evidence. And across cycles, even the observations that are recorded are not standardised — morphology grading is subjective and timing conventions vary between operators, making records difficult to aggregate meaningfully. Without consistent, structured data accumulating over time, the foundation needed for meaningful research or predictive modeling never forms. The Vienna Consensus called for systematic, transparent, and consistent data collection and KPI tracking across IVF laboratories1. The headline KPIs are achievable in a well-run manual lab, but the process-level data layer that would let the field move from KPI reporting to mechanistic understanding remains out of reach.

Conceivable's Approach: The C:NEXUS Platform

Conceivable’s response to that gap is architectural rather than incremental. AURA is a suite of robotic instruments, each designed to provide automation assistance for one step of the IVF workflow; dispensing, imaging, labelling, sperm and egg preparation, fertilisation, culture, and vitrification. C:NEXUS is the software layer that connects them. A central order manager tracks every specimen, routes it through the correct sequence, and logs every action taken at every step, on every instrument, across the full workflow, including the bench-side handling steps that fall outside the incubator and have historically gone unrecorded.

The resulting difference in data density is substantial: a C:NEXUS-managed cycle generates on the order of 5,000 data points, rising to 8,500+ as full station integration completes, and captures structured outputs from sixteen computational models running across the workflow. This produces a structured record covering the procedural steps, environmental conditions, timing intervals, and imaging that a manual workflow captures only in fragments. The Electronic Laboratory Notebook assembles this into a case record as the cycle runs and timestamps the compliance steps alongside it. The record becomes a product of the workflow itself, the only configuration that guarantees documentation survives time pressure. The Vienna Consensus described what systematic laboratory data should look like; this is what it looks like when the workflow generates it natively.

The Implication

For the clinical team, the immediate change is diagnostic: a comprehensive process record to interrogate, one where every procedural step is fully traceable and every clinical decision is systematically documented. The implication extends well beyond individual cycle review. IVF research has long been constrained by the same capture limitations that affect clinical practice; studies are underpowered, datasets are heterogeneous, and technical variability between operators and sites confounds any attempt to isolate biological signal from procedural noise2. Standardised, continuously captured data across AURA laboratories changes this fundamentally. Multi-center studies become feasible with genuine experimental control. KPIs become comparable across sites. The field gains the consistent, structured datasets it has always needed to answer foundational questions about embryo development, culture conditions, and implantation, rigorously, at scale.

And looking further forward: the dataset being built today is the foundation for what becomes possible tomorrow. Predictive models require volume, consistency, and structure. The IVF laboratory that captures everything now is the one best positioned to personalise treatment, anticipate outcomes, and improve results as the science matures. Data density is both an operational advantage and a strategic one that it compounds over time.

The Answer to the Quiz:

A patient’s cycle produces unexpectedly few blastocysts. The clinical team needs to understand why. Which would be more useful to have in front of them?

OR

A. The embryologist’s recollection of the cycle, the standard morphology grades, and handwritten notes on the lab worksheet.

B. A complete, timestamped record of every procedural step, environmental condition, and image captured across the cycle. Structured the same way as every other cycle the lab has run.

The answer is B. It’s the obvious choice, which is why it’s worth asking what it would take to actually have it. Most IVF laboratories today operate closer to A than they would like. Time-lapse, electronic witnessing systems, and quality management software have improved what gets captured, but these are not necessarily universally available and records remain fragmented across systems. The handling, the timing, the technique, are still largely held in operator memory rather than in a durable record. The consequence is a field that can report its KPIs but cannot easily interrogate them.

Conceivable’s C:NEXUS platform produces option B as a product of how the workflow runs, not as a parallel document. The point is not the volume of data, though the volume is substantial. The point is that it creates the structural conditions for systematic, transparent, consistent data, accumulating in a form that supports both clinical decisions today and the research that will improve them tomorrow.

Nature has its own method for selecting the strongest sperm: swimming. Only motile sperm reach the egg. Everything else falls away. In most manual IVF labs, the process of sperm selection by swimming is replaced by a process that uses a centrifuge, where the whole ejaculate is spun at force. The spin damages healthy sperm before the embryologist ever touches the sample.

By the time a sperm is picked for injection, the harm is already done at the molecular level — and no selection under a microscope can see it. Automation-assisted sperm preparation removes the centrifuge from the equation: motile healthy sperm gently migrate, leaving contaminants behind by design.

The Problem : The Method Damages What It Is Meant to Protect

Sperm preparation in IVF relies on two techniques developed decades ago. Density gradient centrifugation pours the ejaculate onto layers of increasingly dense liquid and spins it — healthy sperm, dead cells, leukocytes, and debris all sinking at different rates, with the sperm fraction collecting at the bottom. Swim-up allows motile sperm to swim upward into clean medium, but only after the sample has already been centrifuged once. The damage precedes the selection.

When sperm is compressed under centrifugal force, leukocytes and dysfunctional sperm release reactive oxygen species — free radicals that attack the membranes of the healthy sperm beside them, stripping the lipids that protect the cell, breaking the DNA strands inside it, and degrading the mitochondria that power its movement [1]. The causality runs both ways: the dysfunctional cells cause damage, and the mechanical stress of centrifugation itself further impairs sperm function. The sperm that survives looks normal under a microscope. Its damage is invisible — and no selection at the injection dish can find it, because the harm is already done.

Centrifugation also fails to clear the preparation of biological contaminants. In the creation of the pellet, bacteria, immature germ cells, and inflammatory cells are compressed alongside healthy sperm and remain in contact throughout a total preparation process of around 60 minutes [2]. Their concentration is reduced. They are not eliminated. Microbiological testing of semen samples confirms bacterial contamination is present in the vast majority of ejaculates before preparation — and persists in the processed fraction after centrifugation [3].

The Vienna Consensus — the authoritative performance standard for IVF laboratories — defines what good sperm preparation looks like: in a quality sample prepared by a skilled embryologist, ≥90% of recovered sperm should show progressive motility, with ≥95% as the aspirational benchmark [4]. These are the standards the best centrifugation-based labs reach on their best days. That means even at benchmark, up to 5% of the recovered fraction — the sperm that made it through preparation — are not progressively motile. Those cells must be identified and discarded manually at the injection dish, one sperm at a time, before any injection begins.

A better method has existed in the literature since 1995. The biological principle is straightforward: motile sperm self-select by swimming through dish-based environments, self-washing free of seminal contaminants in the process [5, 6]. Direct comparative analysis has confirmed that this approach yields superior morphology, membrane integrity, vitality, and nuclear chromatin integrity compared to centrifugation-based methods [7]. What has kept it off the clinical floor is execution: preparing a network of precisely connected microdrops, calibrated in real time to each patient’s sperm concentration and motility, is too labour-intensive and operator-dependent to run reliably across a full clinical day [8]

Conceivable’s Approach : Swim, Not Spin

At Conceivable, we eliminated the centrifuge entirely. C:SPERM, the sperm preparation workstation within AURA, our automation-assisted IVF laboratory, works on a simple principle: motile sperm will swim forward through connected drops of culture medium under their own power; everything that cannot swim will not. The system loads a small amount of ejaculate into a starting drop, links it through a series of medium-filled channels to a clean collection drop, and uses a camera to watch in real time as sperm migrate across. When enough motile sperm have arrived in the collection drop, the system cuts the connection. What remains in the collection drop is a population of sperm cells selected entirely by the ability to swim — the same criterion nature uses — with no mechanical force applied. C:SPERM eliminates contamination entirely. Microbiological testing confirmed complete absence of bacterial and pathogen contamination in the collection fraction — including in samples that tested positive for sexual pathogens before preparation [3]

The clinical evidence for this approach is independently validated. A randomized controlled trial of 1,034 ICSI cycles comparing dish-based horizontal migration against conventional swim-up found cleavage rates of 98.2% versus 92.1% (p=0.0003) and blastocyst rates of 48.1% versus 41.2% (p=0.001) in favor of the centrifuge-free method, with no significant difference in fertilization rate [2]. A multicentric study comparing dish-based preparation directly against density gradient centrifugation found blastocystrates of 58.5% versus 41.7% (p=0.009) and clinical pregnancy rates of 41.9% versus 25.8% (p=0.045) [9]. A 2025 systematic review and meta-analysis of centrifuge-free microfluidic sperm selection, which shares C:SPERM's core principle of motility-based selection without centrifugation, confirmed significantly reduced sperm DNA fragmentation and improved reproductive outcomes versus standard methods across multiple independent datasets [10].

Our own clinical data confirms what the principle predicts. In a pilot study of C:SPERM across six ICSI cycles, biochemical testing confirmed that the sperm collected for injection were fully free of seminal plasma — the fluid that carries the contaminants and oxidative material centrifugation fails to eliminate. Progressive motility was 100% in the automation-assisted group, compared to 88.5% with manual density gradient preparation (p=0.06) — above both the Vienna Consensus competency threshold of ≥90% and the aspirational benchmark of ≥95%. Preparation time was 25 minutes versus 60 minutes manually. Seven single blastocyst transfers produced four ongoing pregnancies and three livebirths — the first babies born after automation-assisted dish-based sperm preparation [11]

The Implication

Fertilization rate is stable across preparation methods. What changes when the centrifuge is removed is everything downstream — cleavage, blastocyst development, pregnancy — the stages affected by sperm DNA integrity rather than motility alone [2, 3]. A prospective randomized trial of centrifuge-free microfluidic sperm preparation found improved blastocyst rates and a higher proportion of euploid embryos compared to density gradient centrifugation [12] — connecting sperm preparation directly to embryo chromosomal integrity, and to the outcomes that matter most at transfer. The patients who stand to gain most are those where that integrity is already compromised: male-factor infertility, elevated DNA fragmentation, poor previous cycles. These are the cases where centrifugation-induced damage compounds an existing vulnerability. For samples with very low sperm concentration or severely compromised motility, centrifugation-based preparation remains necessary — a limitation Conceivable is actively working to address through automation-assisted approaches tailored to these cases.

The centrifuge has been the default not because it is the best method for sperm — but because it was practical. C:SPERM changes that: the preparation that protects sperm is now the preparation that scales.

The Answer to the Quiz:

Which sperm preparation best method preserves sperm motility and eliminates contaminants?

A. Centrifugation-based protocols, where sperm is spun at 2-3 cycles at 300-500g

OR

✅ B. Centrifuge-free dish-based separation, where sperm select themselves by swimming, leaving everything else behind

Centrifuge-free preparation can outperform centrifugation on blastocyst development, pregnancy rates, and sperm DNA integrity — outcomes with real clinical impact. The damage centrifugation causes happens before selection begins, and no microscopic assessment can find it afterward. Removing the centrifuge means the sperm arriving at injection is genuinely healthier, not just apparently so.

However, when sperm concentration or motility is critically low, there simply aren’t enough swimming sperm to select from — and centrifugation remains a necessary tool. For most cases, though, the evidence is clear: swim, don’t spin.

References

1. Aitken RJ, Clarkson JS. Significance of reactive oxygen species and antioxidants in defining the efficacy of sperm preparation techniques. J Androl. 1988;9(6):367–376.

2. Baldini D, Baldini A, Silvestris E, et al. A fast and safe technique for sperm preparation in ICSI treatments within a randomized controlled trial (RCT). Reprod Biol Endocrinol. 2020;18:88.

3. Palini S, Primiterra M, De Stefani S, et al. A new micro swim-up procedure for sperm preparation in ICSI treatments: preliminary microbiological testing. JBRA Assist Reprod. 2016;20(3):94–98.

4. ESHRE Special Interest Group of Embryology and Alpha Scientists in Reproductive Medicine. The Vienna consensus: report of an expert meeting on the development of ART laboratory performance indicators. Reprod BioMed Online. 2017;35(5):494–510.

5. Gordon JW, Chen HL. Penetration of hamster oocytes by human sperm in an in vitro fertilization microchamber after insemination with unprocessed semen. Fertil Steril. 1995;64(2):437–440.

6. Hossain AM, Barik S, Rizk B, et al. Analysis of in vitro migration patterns of human spermatozoa by a Petri dish-based horizontal column. Biol Reprod. 1999;61(2):406–410.

7. Hinting A, Lunardhi H. Better sperm selection for intracytoplasmic sperm injection with the side migration technique. Andrologia. 2001;33(6):343–346.

8. Baldini D, Ferri D, Baldini GM, et al. Sperm selection for ICSI: Do we have a winner? Cells. 2021;10(12):3566.

9. Palini S, De Stefani S, Primiterra M, et al. Comparison of in vitro fertilization outcomes in ICSI cycles after human sperm preparation by density gradient centrifugation and direct micro swim-up without centrifugation. JBRA Assist Reprod. 2017;21(2):89–93.

10. Gisbert Iranzo A, Cano-Extremera M, Hervás I, et al. Sperm selection using microfluidic techniques significantly decreases sperm DNA fragmentation, enhancing reproductive outcomes: a systematic review and meta-analysis. Biology. 2025;14(7):792.

11. Acosta Gomez FJ, Aleriano K, Barragan CP, et al. Pilot clinical evaluation of an automatic horizontal sperm swim-out method for semen preparation. Fertil Steril. 2025;124(Suppl). [Abstract P-280]

12. Godiwala P, Kwieraga J, Almanza E, et al. The impact of microfluidics sperm processing on blastocyst euploidy rates compared with density gradient centrifugation: a sibling oocyte double-blinded prospective randomized clinical trial. Fertil Steril. 2024;122(1):85–94.