Company Spotlight: Volta Medical Volta AF-Xplorer II - AI for Atrial Mapping

Maya: Imagine an algorithm analyzing electrical signals from inside a beating human heart at a sampling rate of 1000Hz, taking just 1.5 seconds of recordings to pinpoint complex spatial anomalies during a procedure.

Sam: Today we are looking exactly at that. We are unpacking the FDA 510(k) summary for the Volta AF-Xplorer II, published by the U.S. Food and Drug Administration on September 12, 2025. And as we always say right at the top, our voices are AI-generated, and this is a summary of a public document.

Maya: Since this is a Company Spotlight episode, we need to be crystal clear. This episode is not sponsored. It must not sound like an advert because it isn't one. We are looking strictly at the public 510(k) summary K251518 from the FDA, not at company marketing material, pitch decks, or press releases.

Sam: Exactly. We are going to cover what this device is cleared to do, how it physically integrates into a hospital setting, the predicate device it was compared with, the performance testing data provided, and what limitations or missing information you should be aware of. Let's start with the basics. What exactly is the Volta AF-Xplorer II, and who is it for?

Maya: The FDA classifies it under the regulation name Programmable Diagnostic Computer, with the product code DQK. It is a Class II medical device. The document states its intended use is to assist operators in the real-time manual or automatic annotation of 3D anatomical and electrical maps of human atria. It specifically looks for the presence of multipolar intra-cardiac atrial electrograms exhibiting spatiotemporal dispersion.

Sam: And it is specifically indicated for use during atrial fibrillation or atrial tachycardia. The target users listed are electrophysiologists, medical personnel like biomedical engineers or nurses, and hospital administrators. It is indicated for adults only. So, mechanically, how does this actually work in a cath lab?

Maya: It is a non-sterile reusable medical device. Mechanically, it is composed of a computing platform and a software application. The computer and its attached display have to be located outside the sterile operating room area. The summary actually gives us the exact hardware specs required to run this algorithm.

Sam: I love when they include the hardware specs. Bringing AI into an operating room isn't just about downloading an app from the cloud. You need local processing power. What does the computer need to have?

Maya: It requires a computer with an Intel Core i7-7700 CPU with an 8MB Cache running up to 4.20 GHz, and 32 GB of RAM. It also needs a TPM, or Trusted Platform Module. For the operating system, it runs on a Debian-based Linux OS, specifically Ubuntu version 22.04LTS.

Sam: Okay, so a robust Linux machine sitting outside the sterile field. How does that computer actually talk to the recording equipment measuring the heart? The document mentions two different types of communication setups depending on what systems the hospital already has.

Maya: Right. The first type is unidirectional analog communication. This is used to connect to EP recording systems via a custom-made cable, which the summary notes comes in two variants: DSUB or Octopus. That cable connects to an Advantech PCI-1713U analog-to-digital converter card integrated into the computer, which acquires the analog data, digitizes it, and transmits it to the software.

Sam: And the second type is bidirectional digital communication. This uses an ethernet cable. They even specify the cable comes in four different lengths: 20, 10, 5, or 2m. This ethernet connection transmits the digital signals directly to the computer, and importantly, it allows the systems to talk back and forth.

Maya: Exactly. But a hospital administrator can't just plug this into any random mapping system. The document provides a very specific list of compatible acquisition systems. For EP recording systems, it works with the LabSystem Pro EP Recording System by Boston Scientific, and the MacLab CardioLab EP Recording System by General Electric.

Sam: And for 3D mapping, it works with the EnSite X 3D mapping system by Abbott. So, Boston Scientific, GE, and Abbott systems are supported. What about the actual catheters that go into the patient's heart to record these signals?

Maya: The summary states it works with all existing 510(k)-cleared catheters that meet specific dimension requirements, and it provides a precise list. It supports the Pentaray Nav, the Octaray 2-2-2, Octaray 2-5-2, Octaray 3-3-3, and the Lasso 20, all from Biosense Webster. From Abbott, it supports the Advisor HD Grid and the Advisor HD Grid X. And from Boston Scientific, it supports the Intellamap Orion.

Sam: This leads us nicely to the predicate device. For our listeners, a 510(k) clearance is based on substantial equivalence. It means the FDA agrees a new device is equivalent to a legally marketed predicate. You don't have to run massive new clinical trials if you can prove your device is just as safe and effective. What is the predicate here?

Maya: The predicate device is the original Volta AF-Xplorer, cleared under submission K243812. That is Volta Medical's previous version of this same system. So, the FDA's job here was evaluating what changed between version one and version two, and whether those changes raised any new questions of safety or effectiveness.

Sam: So, let's break down those changes. Starting with the catheters you just listed, there are a few very specific operational changes that a cath lab director would absolutely need to know before buying this.

Maya: Yes. The Volta AF-Xplorer II added support for the Advisor HD Grid X catheter, which replaces the Advisor HD Grid. However, it dropped support for the Reflexion HD catheter. Also, the predicate device supported "any circular 20 poles catheter". The new device does not support that anymore; the Lasso 20 is the only 20 poles catheter that remains compatible.

Sam: That is a major detail. If your lab relies on a different circular 20 poles catheter, this new software version will not support it according to this document. The FDA notes that these limited changes do not affect the safety or performance of the device, but operationally, it changes exactly what you can pull off the shelf.

Maya: Absolutely. Another operational change is the operating system. It was upgraded from Ubuntu version 20.04LTS in the predicate to version 22.04LTS in the new device. And they added the ability to connect to a specific EP recording system through bidirectional digital communication, whereas previously that was only available through analog communication.

Sam: What about the AI algorithm itself? Did they fundamentally change how it detects this spatiotemporal dispersion in the heart?

Maya: The fundamental principles are the same. It still relies on combining signal processing and machine-generated features into a single output. It still analyzes 1.5 seconds of electrogram recordings at a sampling rate of 1000Hz. However, the summary states the architecture and implementation have been updated to include more "state of the art classifiers".

Sam: State of the art classifiers is a nice regulatory phrase, but what data did they actually use to train these new classifiers? Do they provide the specific details of the data set?

Maya: They provide the categories of data that were added, but they completely omit the exact numbers. The document says the dataset used to train the algorithm has been extended compared to the predicate device. They added more intracardiac electrogram samples, a higher number of ablation procedures, and more experts.

Sam: But importantly, the summary does not tell us exactly how many procedures, how many patients, or how many samples were used. It simply states that only the volume of the dataset was changed, not the type of data or the manner in which it was captured.

Maya: That is a crucial limitation to point out for anyone evaluating this tool. As a buyer or a clinician, the 510(k) summary tells you the dataset is bigger than the previous version, but it does not give you the raw numbers to evaluate the scale or diversity of that training data yourself.

Sam: Right. But we do get a lot of detail on how the output is presented to the user, and this interface has changed significantly. In the previous version, the system showed two levels of dispersion: an orange tag for a high likelihood, and a red tag for a very high likelihood of dispersion.

Maya: The summary notes the subject device now groups the previous categories of high and very high likelihood into a single high likelihood category, which uses only the color red. This update does not alter how the device is used because operators previously mapped a dispersion at that location regardless of the specific level.

Sam: Simplifying the visual interface makes sense in a high-stress environment like an ablation procedure where you don't want to be second-guessing shades of color. But they also added a new visual feature, didn't they?

Maya: They did. They added an extra analysis feature. This acts as a subgroup for the red tags, evaluating how consistent the dispersion is over time and intensity for each bipole. It is visually displayed on screen as a gauge that slowly gets filled up.

Sam: A gauge that slowly gets filled up. That gives the electrophysiologist a real-time sense of stability for that specific signal over time. The document mentions this extra analysis moves toward a two-level hierarchy of regions of interest. So how does a doctor actually interact with these tags on the 3D map?

Maya: The identified regions of interest can be manually tagged by the operator in all configurations. But, if the system is using bidirectional digital communication with the EnSite X 3D mapping system, the validated dispersion areas can be automatically displayed in the 3D mapping system as tags in the 3D atrial shell.

Sam: Automatic tagging sounds like a massive workflow upgrade, assuming it is accurate and the tags go exactly where they are supposed to. Which brings us to the performance testing. How did Volta Medical prove to the FDA that this new algorithm, with its new classifiers and visual gauges, actually performs properly?

Maya: They provided two main pieces of performance data in the summary to prove substantial equivalence. First, they performed unitary tests on each algorithm unit. The result was a conformity rate of 100%. This confirmed that the expected performances of the algorithm units were reached.

Sam: A 100% conformity rate is a perfect score on a software bench test. But what about testing it in a clinical scenario? You can't just run unit tests on code and call it a day when it comes to mapping a human heart.

Maya: They also conducted bench testing that included dispersion detection validation by expert assessment, comparing it directly to the predicate device. They verified that the dispersion maps from the Volta AF-Xplorer II using the autotagging functionality were demonstrated to be satisfactory for the clinician, and did not significantly deviate from the manual maps.

Sam: And how was that confirmed? Was there a statistical analysis of the maps?

Maya: It was confirmed by qualitative analysis of the 2D maps. They also tested the new visual gauge that fills up. The highly stable dispersion tags, which they refer to as a "simple or smart de-densification of dispersion", were validated by experts through quantitative and qualitative analysis of 2D dispersion maps.

Sam: And where were those 2D dispersion maps generated? The document specifically notes they were generated during ablation procedures in a simulated environment.

Maya: Exactly. The testing was done in a simulated environment. The FDA summary does not contain data from a prospective, randomized clinical trial showing patient outcomes like long-term freedom from atrial fibrillation. The testing here is based on expert validation of 2D maps in a simulated setting to prove it is as safe and effective as the predicate device.

Sam: This is the nature of the 510(k) pathway. If you prove your new version does the same thing as the old version, you don't need a new massive trial. The FDA’s conclusion is that the minor technological differences raise no new issues of safety or effectiveness.

Maya: And the FDA's clearance letter to Volta Medical's regulatory contact comes with standard warnings. The letter explicitly states that a substantial equivalence determination does not mean the FDA has determined the device complies with all other Federal statutes.

Sam: Yes, the letter reminds the manufacturer that they must comply with general controls like the Quality System regulation, which covers design controls and corrective and preventive action. They also have to comply with the Unique Device Identification System rule, meaning the device needs a UDI on its label and package.

Maya: So, let's pull this all together for a crisp recap. If you are a hospital system looking at AI for your electrophysiology lab, Volta Medical has secured clearance for an updated platform. It relies on a specific computing hardware setup just outside the sterile field, analyzing signals at 1000Hz.

Sam: It offers automatic tagging if you have the Abbott EnSite X system, but you need to pay very close attention to their exact list of compatible catheters and recording systems, as some support from the previous version has been removed. The AI relies on an updated algorithm trained on an extended, though numerically unspecified, dataset.

Maya: And the interface now uses a simplified red tag for high likelihood of dispersion, alongside a new gauge that fills up to indicate stability. The FDA clearance is supported by unitary testing with a 100% conformity rate and expert assessment of 2D maps in a simulated environment.

Sam: That wraps up our Company Spotlight on the Volta AF-Xplorer II. We have linked the full FDA 510(k) summary document in the show notes so you can read the exact hardware specs and catheter lists for yourself. And as a reminder, this podcast is for informational purposes only and is not medical advice.