A surgeon doing a robotic prostatectomy makes one call on each side of the gland: spare the nerve bundle that runs along the prostate, or take it out with the specimen. Spare it and the patient has a better chance of keeping continence and sexual function. Spare it where the cancer has already grown through the capsule, and the surgeon may leave tumor at the edge of the cut.
Pathologists report that leftover tumor as a positive margin. On October 20, Avenda Health announced Surgery Mode, which puts its Unfold AI cancer map in front of the surgeon while the nerve-sparing call is being made. The launch cites research predicting the software could cut positive margins in half. That research never put the map in front of a surgeon.
What Surgery Mode adds
Unfold AI has held FDA 510(k) clearance since December 2022, when it was still called iQuest. It combines the patient’s MRI, the location and pathology of every biopsy core, and the PSA level, then builds a 3D map of where the cancer probably sits. Until now urologists used it before surgery, for counseling and planning. Surgery Mode, as The Robot Report describes it, brings that map into the operating room: larger 3D views, the MRI slices underneath, and a risk score for each quadrant of the capsule, oriented to match what the surgeon sees on the robot’s console. The announcement doesn’t say whether the new mode needed its own FDA submission.
What the study measured
The paper the announcement links is by Priester and colleagues, in BJUI Compass in 2024. The team went back through the records of 147 UCLA patients who had an MRI-targeted biopsy and then had the prostate removed between 2016 and 2022. They ran each patient’s data through Unfold AI after the fact. Then they checked its estimate of whether cancer had grown through the capsule against what a pathologist found when the removed gland was sliced whole and examined. In the back half of the gland, where the nerve bundles run, Unfold AI missed that growth in 16% of the quadrants where it was present. MRI missed it in 27%. The map also beat a PSMA PET scan and the standard risk calculators surgeons use today.
The study measured whether the map agreed with the pathologist. The authors are plain about the rest in their limitations section: retrospective data can’t predict the true clinical impact, and a prospective study is being planned.
The study the claim needs
Evidence for clinical AI answers three questions, and each one takes its own study. Whether the output matches ground truth can be settled with a retrospective dataset and pathology. Whether a surgeon who sees the output decides differently needs the map in the room during real cases. Whether patients end up better off, with fewer positive margins, less recurrence and better continence, needs follow-up after those cases. Avenda’s published study answers the first question, and the margin claim is a claim about patient outcomes.
For a founder building image-based AI for procedures, the first study is the cheap one, because the scans and pathology reports already sit in hospital archives. Clearance for this kind of software, in my experience, usually rests on accuracy against a reference standard. The second and third studies need surgeons, consent forms, follow-up and real money, and they take years to read out.
Hospitals already make that demand of robotic surgery itself. Earlier this month Intuitive’s US da Vinci growth slowed to 10%, and hospitals are asking for evidence that robotic surgery beats laparoscopy. A hospital weighing an AI overlay for those same procedures will ask what patients gained, and a retrospective accuracy study can’t answer it.
Dave’s take
Avenda’s paper states its limits plainly, and the launch would be stronger quoting the result the paper measured. I would fund the prospective study ahead of the next feature. It is the study that can turn a prediction about margins into a result a hospital will pay for.
From Dave’s video library
Dave checks the promise that AI scribes give doctors their evenings back against the published studies, which found about 16 minutes saved per eight-hour shift.
I’m here to help you scale.
Work With DavePrefer a smaller first step? Book a $500 one-hour working session →
Dave Saunders is the founder of Base Reality Group and a Fractional CPO for product companies. He was a founder and operator at Galen Robotics, where the surgical-robotics platform earned FDA De Novo authorization in 2023, and he managed a 35-patent portfolio licensed from Johns Hopkins. He wrote Founders Who Finish and publishes The Build. More about Dave →