Company Spotlight: Iterative Health SKOUT - Live polyp detection with 0ms delay

Sam: Did you know there is now a cleared software tool that draws a blue box around suspected polyps during a live colonoscopy, and the measured video delay it adds to the screen is exactly 0ms?

Maya: It is a fascinating technical claim, and it is right there in the paperwork. Today we are looking at the FDA 510(k) summary for the SKOUT system, published by the U.S. Food and Drug Administration on December 22, 2025, and submitted by Iterative Health.

Sam: Before we get into the details of those technical claims, we need to remind everyone that our voices are AI-generated, and this episode is a summary of a publicly available FDA document, not a review of any company's marketing materials.

Maya: Exactly. We are strictly looking at what is written in the 510(k) summary, document number K253664. Let us start with what this product actually is and who it is for. The FDA document defines the SKOUT system as a software device designed to detect potential colorectal polyps in real time.

Sam: And when they say real time, they mean during the actual procedure, right? Which patient population is this cleared for?

Maya: Yes, it operates during colonoscopy examinations. The indications for use specifically state it is an assistant for adult patients who are undergoing colorectal cancer screening or surveillance.

Sam: I noticed the word 'assistant' there. The document seems very careful about how it frames the role of this software.

Maya: It is extremely careful. The summary explicitly states that the SKOUT system is only intended to assist qualified and trained gastroenterologists. The document stresses that the gastroenterologist is responsible for reviewing the suspected areas and confirming the presence or absence of a polyp based on their own medical judgment.

Sam: So, to be absolutely clear for the clinicians listening, this is not a diagnostic tool that makes the final call. It is an extra set of eyes.

Maya: That is correct. The text explicitly warns that SKOUT is not intended to replace a full patient evaluation. It is also not intended to be relied upon to make a primary interpretation of endoscopic procedures, medical diagnosis, or recommendations of treatment or course of action for patients. And one more important clinical limitation: it is indicated for white light colonoscopy only.

Sam: Okay, that sets the clinical boundaries nicely. I am really curious about how this actually looks in practice for the physician in the room. When they are staring at the monitor, how does the system tell them it has spotted something?

Maya: The document describes a very specific visual interface. The primary graphical user interface feature is the polyp detection notification. When the software spots something, it generates a two-dimensional blue rectangular outline directly around the suspected polyps on the live endoscopic video feed.

Sam: Does it beep or make an alarm sound when the blue box appears? I can imagine an audio alert might get annoying during a long procedure.

Maya: The summary notes that specifically. It says SKOUT markers are not accompanied by a sound. It is entirely visual. And if there is no polyp detected, the bounding box simply does not appear.

Sam: That makes sense for the workflow. But what happens when the physician actually goes in to remove the polyp or take a biopsy? You would not want a big blue box covering up the exact tissue you are trying to cut.

Maya: The designers accounted for that. The document explains that the system pauses polyp detection when endoscopic surgical tools are detected in the video feed. The text states this is to ensure that the bounding box does not hinder any surgical procedure, biopsy, or resection. It also says it may pause when lighting conditions are deemed to be inadequate.

Sam: That is a really smart workflow integration. It knows when to get out of the way. How does the doctor know what state the system is in? Like, whether it is actively scanning or paused because a tool is in the way?

Maya: There is a device status indicator in the bottom-right corner of the screen. The summary breaks down four distinct modes. When the device is actively processing video, it shows a concentric green circle within a larger ring and the SKOUT label below. When a surgical tool is present, that changes to a concentric gray solid circle within a larger dashed ring.

Sam: What about the other two modes?

Maya: There is a bypass mode. If the doctor wants to see the unaltered endoscope processor feed, the indicator turns into a concentric amber solid circle intersected with a diagonal slash across. And finally, there is an error mode. If something goes wrong, you get a concentric red solid circle with a diagonal slash, plus a red X error icon with a code and message right in the center of the screen, over a dimmed endoscope feed.

Sam: It sounds like a very highly engineered user interface. But how did this actually get cleared by the FDA? Is this a brand new clearance with a massive new clinical trial?

Maya: No, this is a 510(k) clearance, which means it was cleared based on proving it is substantially equivalent to a predicate device already on the market. In this case, the predicate device is an earlier version of their own product. The predicate is listed as the SKOUT system, Iterative Health, clearance number K251126.

Sam: Ah, okay. So this is version 2.3 of the SKOUT system being compared to version 2.2. If the old version was already cleared, what exactly did they change in this new submission? Why go back to the FDA?

Maya: The summary lists several minor changes to the technological characteristics. The biggest one involves the AI model itself. The text states they did retraining and refinement of inference algorithms for compatibility with Endocuff devices.

Sam: Endocuff devices? The document does not explain what those are, does it?

Maya: No, it does not explain what an Endocuff is. It just states that the algorithms were retrained for optimized performance in videos with Endocuff, as well as for multiple polyps, and for sessile serrated lesions. The stated goal of this retraining was also to achieve a reduced false positive rate.

Sam: It is interesting that they specifically mention sessile serrated lesions. Even without external medical context, the fact that they explicitly called out that specific type of lesion in the software retraining highlights that it must have been an area where they wanted to improve detection.

Maya: Exactly. They also made some hardware and software compatibility tweaks. The summary mentions a refinement of device software for optimized detection of the edges of the active video feed specifically for the Fujifilm Eluxeo VP-7000 video processor. They also improved computer memory efficiency and optimized the tracking of polyps across video frames.

Sam: Speaking of hardware compatibility, if I am a hospital system leader looking to buy this, what equipment do I actually need? Does it work with any camera?

Maya: It is not universal. The document lists exactly three compatible endoscopes: the Olympus EVIS EXERA II, the Olympus EVIS EXERA III, and the FUJI Eluxeo VP-7000, provided they have a compatible HD output.

Sam: And how does it connect? Does it slow down the video feed? Because if a surgeon is moving a scope around, even a tiny delay in the video could be incredibly disorienting.

Maya: The video input uses a Serial Digital Interface, or SDI, with de-interlacing applied during pre-processing. As for the delay, this is one of the most striking numbers in the summary. They conducted bench testing for video delays, and the document reports that the video delay due to the device is 0ms, and the video delay due to marker annotation is also 0ms.

Sam: Wait, literally zero milliseconds? That seems almost impossible for a system actively running an AI inference algorithm on every frame of a high-definition video.

Maya: That is exactly what the summary reports: 0ms. However, it does include a parenthetical note right next to that figure. It says there is no standard error, all results were 0, but it also notes a minimum resolution of 1.1ms. Furthermore, the testing confirmed that no pixel level degradation is introduced by SKOUT to the endoscopic system.

Sam: Okay, so the resolution of their testing equipment bottomed out at 1.1ms, and they didn't detect any delay above that threshold. That is still remarkably fast. It explains why they can claim it works in real time.

Maya: Yes, for the user in the room, it should feel instantaneous. There is also a hardware safety feature mentioned. There is a Mode Selection Button that allows for instantaneous toggling between the SKOUT video feed and the bypass video feed in the event of a software error that affects video quality.

Sam: Let us talk about the performance testing. You mentioned earlier they aimed for a reduced false positive rate. Does the summary give us the actual accuracy numbers? What was the sensitivity and specificity?

Maya: This is where we have to point out the limitations of a 510(k) summary document. It says that standalone and on-device algorithm performance testing was conducted for the evaluation of true positives, false positives, and polyp detection time. It also mentions they used an expanded dataset for the added Endocuff compatibility.

Sam: Right, but what were the results?

Maya: The summary does not provide them. It does not state the size of the dataset, the demographic breakdown of the patients, or the final percentages for true positives and false positives. It simply concludes that the SKOUT system demonstrated passing results in all applicable testing.

Sam: That is a bit frustrating from a transparency perspective, though it is quite common in these public summaries. If they do not provide the clinical data here, how do we know it actually works in a patient?

Maya: The manufacturer argues in the conclusion that because the inference algorithms utilize the same architecture and meet the same performance requirements as the predicate device, the clinical performance remains unchanged from the clinical performance submitted in an even older clearance, which they list as K213686. Because the fundamental technology is the same, they only had to prove substantial equivalence through software verification and non-clinical bench testing.

Sam: So if a hospital wants to see the actual clinical trial data proving this increases adenoma detection rates, they will not find it in this specific December 2025 document. They would have to dig back into older filings or ask the company directly.

Maya: Exactly. The FDA letter attached to the summary also includes standard boilerplate language that buyers should remember. The FDA states that this substantial equivalence determination does not mean that the FDA has made a determination that the device complies with other requirements of the Act or any Federal statutes and regulations. It is a Class II device subject to general controls, like good manufacturing practice and truthful labeling.

Sam: The letter also notes that the FDA's Center for Devices and Radiological Health does not evaluate information related to contract liability warranties. Basically, the FDA is clearing it to be marketed, not guaranteeing your hospital's purchase contract.

Maya: Precisely. But taking a step back, the key takeaway from this document for the medtech community is how AI is moving seamlessly into the live surgical environment. The SKOUT system is designed to process high-definition video, run an inference algorithm, and draw a bounding box over a polyp with a measured delay of 0ms, and it knows to turn itself off the second a biopsy tool enters the frame.

Sam: It is a clear shift from AI that just analyzes images after the fact, to AI that acts as a real-time copilot in the endoscopy suite, without disrupting the physician's established workflow.

Maya: That is the promise outlined in the clearance. But again, it remains an assistant, not a replacement for the gastroenterologist's own medical judgment.

Sam: That is a great place to wrap up. We have linked the full FDA 510(k) summary for the Iterative Health SKOUT system in the show notes so you can read the exact technical specifications yourself. As always, this podcast is for informational purposes and is not medical advice. Thanks for joining us on this deep dive.