Virtual Balloons, Real Hearts: Simulations Now Predict a Key TAVI Correction Step

When a transcatheter aortic valve fails to open fully inside a diseased heart, surgeons sometimes reach for a small, powerful solution: a balloon. A new study published in the Annals of Biomedical Engineering has, for the first time, built and rigorously validated a computer model of that corrective step, known as balloon post-dilation, inside patient-specific simulations of transcatheter aortic valve implantation, or TAVI. The work, led by researchers at the Politecnico di Milano together with clinicians at Humanitas Research Hospital and Fondazione IRCCS Cà Granda in Milan, shows that adding this often-ignored procedural phase to virtual heart models slashes prediction errors by as much as 80 percent, potentially paving the way toward digital rehearsal of heart valve procedures before a single incision is made.

TAVI has transformed the treatment of aortic stenosis, the narrowing of the heart’s main outflow valve that affects millions of older adults worldwide. Instead of opening the chest, clinicians thread a collapsible biological valve through an artery and deploy it inside the native valve’s calcified frame. Yet the procedure is not always perfect. In a substantial fraction of patients, the new valve does not expand symmetrically against the hardened calcium of the aortic root, leaving gaps through which blood leaks backward, a complication called paravalvular leakage. Mild to moderate leakage affects between 7 and 40 percent of TAVI patients, and when moderate or severe it is strongly associated with increased late mortality. The standard remedy is balloon post-dilation, in which a non-compliant balloon is inflated inside the freshly implanted stent to push calcifications outward and force the valve frame into fuller contact with the vessel wall. It is performed in roughly 13 to 54 percent of cases depending on the registry.

Despite how common the maneuver is, computational models of TAVI have almost universally ignored it. A recent scoping review cited by the authors examined 40 studies that used numerical simulation to predict TAVI outcomes and found that not a single one modeled balloon post-dilation. Two prior efforts, by Li and colleagues and by Akodad and colleagues, attempted to simulate the ballooning step, but neither calibrated the balloon’s material properties through physical testing nor validated their predictions against real post-operative images. That gap matters because the decision to perform post-dilation is a delicate risk-benefit calculation: the maneuver can enlarge the valve area, lower pressure gradients, and reduce leakage, but it carries risks including annular rupture, valve migration, stroke, conduction disturbances, and damage to the delicate leaflets of the new valve.

To close that gap, the Milan-based team reconstructed a True Dilatation non-compliant balloon, a commercially available device from BD, using reverse engineering. They photographed a physical balloon at high resolution, measured its dimensions with a digital caliper, and processed the images in ImageJ to extract precise geometry, then built a computer-aided design model reproducing the device’s 85-millimeter total length, 45-millimeter working length, and 21-millimeter maximum diameter, later scaling it to the 22, 24, and 26 millimeter sizes actually used in patients. The balloon wall was represented as a shell just 0.3 millimeters thick, and the model was meshed with triangular shell elements averaging 0.8 millimeters, matching the refinement of the stent mesh to keep contact calculations stable.

Crucially, the team did not simply assume the balloon’s mechanical behavior; they measured it. Specimens cut from real balloons were stretched in a bioreactor equipped with a precision load cell, in both the longitudinal and circumferential directions, through six full loading-unloading cycles. The resulting stress-strain curves revealed a Young’s modulus of approximately 600 megapascals, with nearly identical values in both directions, indicating that the material behaves as an essentially isotropic elastic solid. To validate this finding, the researchers inflated an intact 21-millimeter balloon with water in a custom 3D-printed rig, raising the pressure in half-atmosphere steps up to four atmospheres while measuring the external diameter with a caliper. When they replicated this experiment in the virtual world using their 600-megapascal material model, the simulated pressure-diameter curve matched the experimental measurements with a coefficient of determination of 0.97, with the closest agreement near the nominal working pressure of three atmospheres.

With the balloon model calibrated and validated, the researchers integrated it into the final configurations of patient-specific TAVI simulations previously developed by the same group. Four patients who underwent TAVI at Humanitas Research Hospital between 2017 and 2024 were retrospectively included, all of whom had received self-expanding valves and post-dilation, and all of whom had post-operative CT scans available for comparison. In the simulations, the crimped balloon was placed inside the virtually implanted stent, inflated to its nominal pressure to push calcifications against the aortic wall and fully expand the Nitinol frame, then deflated to capture any elastic recoil. Two of the four patients also carried mechanical mitral valve prostheses, which the models explicitly represented as rigid bodies to capture potential device-device interference. All computations ran on an explicit finite element solver, LS-DYNA, using damping and selective mass scaling to keep energy ratios under five percent and maintain a constant time increment of one microsecond.

The validation against clinical imaging delivered striking results. Comparing the simulated final stent geometry with that segmented from post-operative CT scans, the average error in orifice area across three cross-sectional planes was just 1.55 percent, with a standard deviation of 1.26 percent, while the error in stent eccentricity at the level of the left ventricular outflow tract averaged 1.54. More telling was the comparison with simulations that omitted post-dilation: including the ballooning step reduced the orifice area error from 3.1 percent to 1.6 percent, a 53 percent improvement, and slashed the eccentricity error from 7.4 percent to 1.5 percent, an 80 percent improvement. Centerline analysis, which measures how closely the simulated stent axis follows the real one, found a mean deviation of only 0.386 millimeters across all patients.

The researchers also explored what the geometry changes mean physiologically. Using a simplified geometric surrogate for paravalvular leakage based on peri-prosthetic flow volume, they found that post-dilation reduced the leakage-prone space in three of the four patients, suggesting improved sealing between device and wall. The fourth patient, who started with a very low baseline flow volume, showed essentially no change, a reminder that the benefits of ballooning depend heavily on each patient’s unique anatomy. The authors explain that the key mechanism is not material damage but the redistribution of contact constraints: the balloon displaces calcified deposits, which alters the available space for stent expansion, so after deflation the Nitinol frame settles into a new equilibrium that would never be predicted by a deployment-only simulation. Because calcification patterns differ so dramatically between patients, the effects of post-dilation can be non-uniform and even counterintuitive, which is precisely why patient-specific modeling is needed.

The implications reach beyond academic curiosity. The authors argue that a validated virtual post-dilation capability could support pre-operative planning in which clinicians first examine the predicted post-deployment configuration of the valve, checking stent expansion, eccentricity, and apposition to the annulus, and then virtually test whether ballooning with different balloon sizes would deliver meaningful improvement at an acceptable safety margin. Such digital twins could help operators decide when corrective ballooning is likely to help, and when it merely adds risk. The study does carry limitations: only four patients were analyzed, partly because routine post-operative CT imaging is not standard practice, and the models do not simulate aortic wall rupture or calcification fracture, nor do they compute actual blood flow for leakage quantification. Still, the researchers see this work as a clear step toward credible, clinically deployable in silico tools. Notably, a review of the field found that only about 10 percent of TAVI simulation studies validated their results against post-implant CT scans at all. By calibrating every component against benchtop experiments and every prediction against real clinical outcomes, this study sets a new benchmark for what rigorous validation in computational cardiology should look like, and it brings the vision of personalized, simulation-guided structural heart intervention measurably closer to the operating room.

Subject of Research: Patient-specific computational modeling and validation of balloon post-dilation in transcatheter aortic valve implantation

Article Title: In Silico Modeling and Validation of Post-Dilation in TAVI Patients

Article References: Perri, L. M., Grossi, B., Fregona, V., Barati, S., Barbieri, L., Tumminello, G. D., Carugo, S., Bianchi, E., De Stefano, P., Berti, F., Cozzi, O., Condorelli, G., Stefanini, G., Dubini, G., Rodriguez-Matas, J. F., Migliavacca, F., & Luraghi, G. (2026). In Silico Modeling and Validation of Post-Dilation in TAVI Patients. Annals of Biomedical Engineering. https://doi.org/10.1007/s10439-026-04325-0

Image Credits: AI Generated

DOI: 10.1007/s10439-026-04325-0

Keywords: TAVI, balloon post-dilation, finite element analysis, patient-specific simulation, paravalvular leakage, aortic stenosis, Nitinol stent, computed tomography, in silico medicine, biomedical engineering, model validation, digital twins