PULSE: Dataset for Studying Deformation in 3D Patient-Specific Pulmonary Artery Anatomies

Background: The pulmonary valve and right ventricular outflow tract (RVOT) direct blood from the right ventricle to the pulmonary artery (PA). These structures are often affected in congenital heart disease (CHD), which occurs in ~1% of live births. RVOT obstruction and associated conditions, like pulmonary regurgitation or Tetralogy of Fallot (ToF) have increased over time, along with pulmonary valve replacements (VR). Surgical repairs and VR can restore patency but may cause long-term complications. Due to complex geometry and variability, RVOT and PA morphological assessment is challenging. Finite element (FE) simulations are a computational tool to analyze tissue mechanics, strain, and structural responses under various conditions. To contribute to cardiovascular research, we introduce the PULmonary artery Strain Evaluation (PULSE) dataset, a curated resource designed to capture circumferential deformation in the RVOT and PA.

Methods: The dataset is derived from 60 patient-specific three-dimensional anatomies reconstructed from cardiac magnetic resonance (CMR) imaging of patients with ToF who underwent transannular patch repair that served as the basis for FE simulations. For each model, an internal pressure was applied and a wall stiffness value assigned, based on mechanical properties of swine PA tissue. FE simulation results were then validated in four random imaging cases: area changes were measured in the CMR phase contrast images, mimicking the transition from diastole to systole and comparing regional deformation to observed values. Additionally, data from one patient outside the database was used for validation: a 3D reconstruction from the Computed Tomography (CT) images at 30% of the cardiac cycle was use for the simulation, and results were compared to the systolic reconstruction (~70%) using Mimics software (Materialise, Leuven, Belgium).

Results: A total of 771 three-dimensional FE models is provided: 60 baseline (undeformed) and 711 deformed configurations generated under varying internal pressures and tissue stiffness values. This combination enables a systematic exploration of how anatomical variability and mechanical properties influence the deformation of the PA and RVOT. FE simulations showed good agreement with CMR measurements across four patients. Absolute errors ranged from 40 to 124 mm², while relative errors in the tissue deformation simulation varied between 3.1% and 12.1%. Also, comparisons between the systolic CT images and the FE simulation on point-to-point surface analysis showed a maximum error of 3.5 mm in specific regions.

Conclusion: The PULSE dataset has different potential applications. For clinicians and researchers, it could provide a foundation for investigating RVOT and PA mechanics. For medical device developers, the dataset could offer anatomical and biomechanical scenarios to inform the design of transcatheter pulmonary valves. For educators and students, it represents a platform for hypothesis-driven research in cardiovascular biomechanics.

Comparison between CMR-derived and FE-simulated cross-sectional areas in the main pulmonary artery in four patients. Each row (A–D) shows the results for one patient, including CMR imaging, flow and area-time curves over the cardiac cycle (R–R interval), and the simulated 3D arterial geometry under catheterization-derived pressure conditions. The vertical dashed lines indicate the time of peak systolic flow, at which the true area (measured from CMR) was extracted and compared with the simulated area.