Think about the problem an operator faces when threading a catheter into an artery to modulate the nerves that run alongside it. The nerves themselves are invisible on the imaging. What is visible is the vessel — its walls, branches, and the way blood moves through it. The question is where along that vessel the device should actually act, and, just as importantly, where it should not. A patent that issued on June 30, 2026, assigned to Medtronic Ardian Luxembourg S.a.r.l., US12667425B2, is directed squarely at that question, and its answer is to let fluid physics draw the map.

The short version is this: the disclosed system takes a picture of the vessel and a reading of how blood is flowing through it, turns those into a physics simulation, and uses the simulation to mark two kinds of spots — places to treat and places to steer clear of. In the language of the patent, the inputs are three-dimensional imaging of a target blood vessel and corresponding hemodynamic data, and the engine that consumes them is a computational fluid dynamics model. CFD is the same class of simulation that aerospace engineers use to model airflow over a wing; here it is pointed at the flow of blood through a specific patient's vessel.

Independent claim 1 spells out the pieces without reference to any particular disease. It recites a processor configured to receive digital data about at least one feature of a blood vessel, receive hemodynamic data from at least one sensor, generate a model of the vessel from both, determine a target region or an avoidance region based on the model, and output an indication of those regions. The claim is written around the data pipeline — image in, flow in, model built, regions out — rather than around a clinical result.

A system comprising: a processor configured to: receive digital data including information about at least one feature of a blood vessel of a patient; receive hemodynamic data from at least one sensor; generate a model of the blood vessel based at least in part on the digital data and the hemodynamic data; determine at least one of a target region of the blood vessel for delivery of neuromodulation therapy or an avoidance region of the blood vessel for avoiding delivery of neuromodulation therapy based on the model; and generate an output indicative of at least one of the target region or the avoidance region.— Methods and systems for optimizing perivascular neuromodulation therapy using computational fluid dynamics, US12667425B2

What the model actually looks for

The dependent claims are where the engineering shows. The "feature" of the vessel that feeds the model can be a cross-sectional area, a diameter, a volume, or a length, or a structural landmark — a wall, a lumen, a branch, a bifurcation, a carina, an ostium, a taper region, an aneurysm, a calcification, or an intimal deposit. The imaging that supplies those features can come from X-ray, computed tomography, MRI, fluoroscopy, ultrasound, optical coherence tomography, intracardiac echocardiography, or angiography — a deliberately broad list, because the point is the modeling step, not the scanner. The hemodynamic input can be blood pressure, blood flow, blood impedance, or viscosity, sensed either from a sensor on the neuromodulation catheter itself or from an external device.

The most telling limitations concern how the system decides an avoidance region. One dependent claim identifies avoidance zones by looking for portions of the vessel with low or high wall shear stress, high shear-stress gradients, or flow features like separation, eddies, impinged flow, turbulent flow, or secondary flow — plus anatomical features such as an ostium, a carina, a calcification, or an aneurysm. In plain terms, the model is hunting for the places where the fluid physics gets messy or the anatomy is fragile, and telling the operator to keep the device away from them. Target regions, conversely, are found by comparing a hemodynamic parameter at a given location against a threshold value. The system can even take in the live position of the neuromodulation catheter and output a recommendation of whether to proceed at that spot.

What ties this to a real instrument rather than a pure software exercise is that the claim set includes the catheter. The system "further comprising a neuromodulation catheter" places the modeling engine inside a hardware loop: a device goes into the vessel, sensors report flow, the model updates its map, and a user interface displays the vessel with visual markers indicating the target and avoidance regions. It is a planning-and-guidance layer wrapped around a physical catheter.

One grant in a device cluster

The CFD grant did not issue alone. The same June 30 drop carried a run of grants to Medtronic entities that, read together, sketch how the company is instrumenting the space around its neuromodulation and navigation hardware. On the imaging side, US12670646B2, assigned to Medtronic Navigation, is directed to reconstructing an extended field-of-view image from an X-ray gantry by projecting and interpolating padding data between a central region and a surrounding annulus — a way to widen what the imaging system can show without a bigger detector. That is the kind of imaging that feeds a model like the one in the CFD patent.

On the hardware side, US12667719B2 is directed to an implantable lead whose fixation mechanism anchors an electrode array within a foramen of the sacrum to generate a stimulation field, and US12667416B2 is directed to a renal neuromodulation device with an elongated shaft that delivers a thermal element to a renal artery by an intravascular path. Both are the physical business end of neuromodulation — the parts that a planning model like US12667425B2 exists to aim. And on the sensing side, US12667306B2 is directed to a method that reads a patient's local field potential — a bioelectric signal — to infer timing information and output it to a processor, extending the same instrument-plus-data pattern into monitoring.

Seen as a set, the throughline is consistent: pair a device with a data model. An extended-FOV image, a CFD simulation of vessel flow, an implantable electrode array, an intravascular thermal element, and a bioelectric-signal reader are five different pieces, but each grant puts a sensor or an image on one side and a computed output on the other. The CFD patent is the clearest single illustration of the idea, because it is explicitly about converting an image and a flow measurement into a decision surface — a marked-up vessel that tells the operator where the catheter's job is, and where it isn't.

The usual reading caveats apply. These are issued grants, so the claims are fixed as allowed rather than pending, but a patent describes what is claimed, not what has shipped or how well any of it performs. The CFD grant is directed to the modeling-and-guidance system itself; it does not, and this article does not, make any claim about clinical results. What it does show, cleanly, is a design philosophy: when the target is invisible, model the thing you can see, and let the model draw the boundaries.