ISOTROPIC 4D WHOLE THORACIC AUTOMATED CARDIAC MRI (AUTOCMR) FOR NON-INVASIVE COMPUTATIONAL SIMULATION OF PATIENT-SPECIFIC LEFT VENTRICULAR DYNAMICS (RF_FR_309)

In patient-specific cardiac simulations, selecting the correct unloaded geometry is essential due to residual stress and strain effects [1]. The process involves segmentation, smoothing, surface triangulation, and optimization to ensure accurate finite element (FE) meshes [2]. Methods such as the Klotz empirical relation [3], inverse elastic problem [4], and inverse displacement approach [5] have been developed to derive unloaded geometry from in vivo data, with studies [6] showing different configurations impact material parameters, stresses, and strains in personalized cardiac models. This study utilizes advanced 4D whole thoracic automated cardiac MRI (AutoCMR) from four patients. High spatiotemporal resolution isotropic imaging reduces the need for pre-processing and interpolation of the segmented data. Compared to conventional CT imaging with limited temporal resolution, AutoCMR captures more cardiac phases, enabling detailed strain and motion analysis. By comparing volumetric strain from simulations with MRI segmentations, we assess left ventricular (LV) dynamics and explore how advanced imaging could enhance the accuracy and efficiency of future computational studies.

Methods:

We conducted FE simulations of the LV using segmentations of 4D patient-specific cardiac geometries obtained from AutoCMR scans from four patients (3F, Mean age: 54.5 ± 18.08 years). AutoCMR uses free-running spoiled gradient echo (GRE) sequences, has a resolution of 1.6 mm × 1.6 mm × 1.6 mm, covering a field of view (FOV) of 307 mm × 307 mm × 307 mm, across 30 cardiac phases. AutoCMR data was captured using a 3T MR system (MAGNETOM Cima.X, Siemens Healthineers, Forchheim, Germany). We segmented the LV using 3D Slicer, processed the geometry with a custom Python script, and generated meshes in Gmsh. Fiber architecture was applied using the Rule-Based algorithm [7] (-60° to 60° orientation). The myocardium was modeled as a transversely isotropic hyperelastic Fung material with active contraction in VUMAT for ABAQUS Explicit simulation, including a closed-loop circulatory system. Active stress and elastic modulus were adjusted for each subject to match end-systolic and end-diastolic volumes.

Results:

Our simulations generated pressure-volume (PV) loops for four subjects using high-resolution LV geometries from 4D AutoCMR, requiring minimal pre-processing. The models were evaluated by comparing simulated volumes across the cardiac cycle with MRI-derived volumes, revealing strong alignment. The Pearson correlation coefficients between segmentation and simulation volumes were consistently above 0.93, demonstrating high agreement across all subjects. A scatter plot comparing simulation and segmentation volumes across all frames showed a linear relationship, with a Pearson correlation coefficient of 0.97. Minor discrepancies in absolute volumes were observed but did not significantly affect the overall trends.

Conclusion:

Further material parameter optimization could improve the clinical relevance of these simulations. Automating the workflow using high-resolution AutoCMR data will streamline the process, allowing clinicians to quickly transition from image acquisition to patient-specific simulations, improving both diagnostic accuracy and treatment planning.