ICL Fit: Pentacam-Based AI Agent for ICL Sizing, Recommending the Best-Fit EVO Visian Implantable Collamer Lens Size for Each Individual Eye
Pentacam-Based AI Agent for ICL Sizing
Best-fit ICL sizing from a Pentacam scan
ICL Fit is an AI ICL sizing tool for EVO and EVO+ Implantable Collamer Lens surgery. It reads an eye’s anterior-segment measurements from a standard Pentacam scan and returns the best-fit ICL size for that eye, so the sizing decision is made from the eye’s internal anatomy rather than inferred from white-to-white alone.
The model is built and maintained by Gurpal Virdi, MD and Matt Hirabayashi, MD, authors of the peer-reviewed VAULT and VAULT-OCT deep-learning studies in the Journal of Cataract & Refractive Surgery, and it learns continuously from real EVO ICL surgical outcomes contributed by the surgeons who use it. The wider body of work behind it is listed in the publications below.
Why ICL sizing is hard
An ICL sits in the ciliary sulcus, and the clearance between the lens and the natural crystalline lens is called the vault. Sitting very close to the crystalline lens raises the risk of anterior subcapsular cataract, and sitting too far forward can crowd the anterior chamber angle. It is tempting to turn that into a target number, but vault is dynamic. Our study of ICL vault and anterior chamber angle under varying lighting conditions measured it shifting within the same eye as lighting changed, and it also moves with accommodation and over the years after surgery, so any single reading is a snapshot. Vault is better read as one visible sign of how well a lens fits an eye. The sizing decision is about overall fit, not about landing a particular figure.
The sulcus diameter that governs fit is not directly visible on most refractive workups, so conventional sizing infers it from white-to-white corneal diameter, which correlates only weakly with the space the lens actually occupies. That is why sizing, not the implant itself, is the step that decides whether a case goes well. We have published on ICL exchanges and explants due to sizing at a high-volume US center, and a single poorly fitting lens is often enough to put a surgeon off the procedure entirely.
How ICL Fit differs from other ICL sizing methods
Most sizing approaches fall into one of three groups:
- Single-variable nomograms such as the STAAR OCOS default, which size primarily from white-to-white and anterior chamber depth.
- Multi-parameter formulas such as the Reinstein, Kojima (KS) and NK nomograms, which add lens rise, pupil size and other anatomic inputs.
- Image-based deep learning, which works from the anterior-segment image itself and generally requires ultrasound biomicroscopy (UBM) or very-high-frequency digital ultrasound to see the sulcus directly.
ICL Fit is none of these. It works from the anterior-segment measurements a Pentacam already produces rather than from images, it learns from real surgical outcomes rather than applying a fixed formula, and it returns the best-fit size for the individual eye.
We built the image-based approach first. VAULT used very-high-frequency ultrasound images and VAULT-OCT used AS-OCT. Both are published, and we moved on from that route deliberately. Sizing from images ties the result to hardware most refractive practices do not own, to a technician trained to acquire reproducible captures, and to the quality of each individual image. Sizing that depends on an acquisition step most clinics cannot perform is not sizing most clinics can use.
Working from measurements removes that dependency. Every ICL candidate can be sized inside the workup a refractive practice already performs, rather than only the patients who get routed to a UBM. The barrier to ICL adoption has never been whether accurate sizing is possible in principle. It is whether it is available on the day, in the room, for the patient in front of you.
Who ICL Fit is for
ICL Fit is clinical decision-support software for ophthalmologists and refractive surgeons performing EVO ICL implantation. It is intended to inform, not replace, the operating surgeon’s sizing judgment, and it is not patient-facing medical advice. Patients considering ICL surgery may find our plain-language overview more useful.
Read more on the science behind vault and sizing in our ICL education notes, the frequently asked questions, or the peer-reviewed publications the model is grounded in.
ICL sizing and vault: common questions
What is a good ICL vault?
A figure of roughly 250 to 750 micrometers is the range most often published, including by STAAR, and it is useful as a broad safety heuristic. Very low vault sits close to the natural crystalline lens and raises the risk of anterior subcapsular cataract. Very high vault can crowd the anterior chamber angle and is associated with pigment dispersion and raised intraocular pressure. But sizing is not a matter of hitting a number, and ICL Fit does not size to a target vault. Vault is dynamic: our own study of ICL vault and anterior chamber angle under varying lighting conditions, published in Clinical Ophthalmology, measured it shifting within the same eye as lighting changed, and it also moves with accommodation and gradually over the years after surgery, so any single reading on anterior-segment OCT is a snapshot rather than a constant, and two eyes sitting at the same vault can be fitted quite differently. What matters is how well the lens fits the individual eye as a whole, because that is what determines how it behaves across real-world conditions.
What sizes does the EVO ICL come in?
The EVO and EVO+ Visian ICL (STAAR Surgical) is made in four overall lengths: 12.1 mm, 12.6 mm, 13.2 mm and 13.7 mm. Because only four sizes exist, sizing is a discrete choice rather than a continuous one, and the same eye can vault very differently on two adjacent sizes. Choosing correctly between them is the central technical problem of ICL surgery.
How is ICL size determined?
Conventional ICL sizing uses white-to-white horizontal corneal diameter and anterior chamber depth fed into a manufacturer nomogram such as OCOS. Anterior chamber depth generally needs to be 3.0 mm or greater for implantation under the US label. Because white-to-white is an external corneal measurement, it is only a proxy for the internal space the lens occupies, so newer methods draw on the fuller set of anterior-segment measurements available from Pentacam Scheimpflug tomography, AS-OCT or ultrasound biomicroscopy, including anterior chamber depth and volume, angle-to-angle distance and crystalline lens rise. The aim is to choose the size that fits the eye best, since fit is what the lens position and behaviour follow from.
What is the difference between white-to-white and sulcus-to-sulcus?
White-to-white is the external horizontal diameter of the cornea, measured limbus to limbus. Sulcus-to-sulcus is the internal diameter of the ciliary sulcus behind the iris, where the ICL haptics seat. White-to-white is an external corneal measurement and correlates only moderately with the internal anatomy the lens actually occupies, which is the structural reason white-to-white nomograms carry inherent sizing error. Sulcus-to-sulcus is closer to the relevant space, but it is one dimension of a three-dimensional fit. Anterior chamber depth and volume, angle configuration and how far the natural crystalline lens sits forward all shape how a given lens will sit, which is why no single diameter, external or internal, settles the sizing question on its own.
Can you size an ICL without a UBM?
Yes. Ultrasound biomicroscopy is valuable because it images the ciliary sulcus directly, but it measures one dimension of a fit that depends on several, and it requires hardware most refractive practices do not own plus a technician trained to acquire reproducible images. Sizing can also be done from the anterior-segment measurements a refractive practice already captures. ICL Fit works from the measurement export of a standard Pentacam scan, covering anterior chamber depth and volume, corneal curvature and the other parameters that describe the space the lens has to occupy, so every ICL candidate can be sized inside an existing workflow rather than only the patients routed to a UBM. It recommends the best-fit size for the individual eye.
Which formulas and nomograms are used for ICL sizing?
Commonly used ICL sizing methods include the STAAR OCOS white-to-white nomogram, the KS formula, the NK formula, the LASSO formulas, the Reinstein formula based on very-high-frequency digital ultrasound, the Kane ICL formula and device-specific regressions such as the ANTERION-based formula. Machine-learning approaches fall into two kinds. Image-based deep learning, including the VAULT and VAULT-OCT models, works from the anterior-segment image itself. Measurement-based models, including ICL Fit, work from the numeric anterior-segment parameters a device already reports, learned from real surgical outcomes rather than fitted as a fixed regression on white-to-white.
What causes a high or low ICL vault?
Vault reflects how a whole lens sits in a whole eye, so it does not trace back to one measurement. Broadly, a lens that is large relative to the space available tends to sit higher and one that is small relative to it tends to sit lower, but that space is not a single dimension. Anterior chamber depth and volume, angle configuration, how far the natural crystalline lens sits forward, ciliary body anatomy and patient age all contribute and interact. This is why two eyes with identical white-to-white, and sometimes even similar sulcus diameters, can behave differently with the same lens. Once implanted, vault continues to vary with pupil size, lighting and accommodation, so it is better understood as the result of overall fit than as the output of any single variable.
How is ICL vault measured after surgery?
Post-operative vault is measured on anterior-segment OCT as the vertical distance between the posterior surface of the ICL and the anterior surface of the natural crystalline lens, read in micrometers. Slit-lamp estimation in corneal-thickness units is a quicker bedside approximation, but OCT is the standard for a precise and repeatable number.
More detail on each of these is in the full FAQ and the ICL education notes.
Peer-reviewed research behind ICL Fit
ICL Fit is built by the team that published the studies below on ICL sizing, fit and outcomes. The full list, with abstracts, is on the publications page.
- Dynamic Changes in Implantable Collamer Lens Vault and Anterior Chamber Angle Under Varying Lighting ConditionsClinical Ophthalmology
- VAULT-OCT: Vault Accuracy Using Deep Learning Technology for Predicting ICL Postoperative Vault with AS-OCTJournal of Cataract & Refractive Surgery
- VAULT: Vault Accuracy Using Deep Learning Technology, a Novel Image-Based AI Model for Predicting ICL Postoperative VaultJournal of Cataract & Refractive Surgery
- ICL Exchanges or Explants Due to Sizing in a United States High Volume CenterClinical Ophthalmology
- Accuracy of Reported Sizes of the EVO/EVO+ Visian Implantable Collamer LensesCureus
- Clinical Outcomes of Implantable Collamer Lenses for the Treatment of Myopia in Eyes with Shallow Anterior Chamber DepthClinical Ophthalmology
- Phakic Intraocular Lenses as a Potential Treatment for Pigment Dispersion SyndromeCureus
ICL Fit is a product of Bimini Vision LLC. EVO, EVO+, Visian, ICL, Collamer and OCOS are trademarks of STAAR Surgical Company. Pentacam is a trademark of OCULUS Optikgeräte GmbH. ICL Fit is not affiliated with, endorsed by, or sponsored by STAAR Surgical Company.