Session: Research Posters

Paper Number: 173315

A Model-Based Approach for Computing Cardiac Deformation From Dense Mri With Minimal User Input 

Accurate assessment of cardiac function is critical for the early diagnosis, monitoring, and treatment of cardiovascular diseases. While conventional clinical metrics such as ejection fraction offer a global view of cardiac performance, they often fail to capture localized tissue dysfunction that may precede organ-level changes. As such, local cardiac strains are a promising tool to provide a more detailed and region-specific understanding of myocardial function throughout the cardiac cycle.  This work focused on developing a pipeline for a robust estimation of cardiac strains using Displacement Encoding with Stimulated Echoes (DENSE) Magnetic Resonance Imaging (MRI), which provides voxel-wise displacement measurements.

Accurate segmentation of the myocardium is a critical prerequisite for the reliable computation of cardiac strains. Misidentification and inclusion of non-myocardial structures can introduce substantial errors into the displacement field and, consequently, the derived deformation measures. For the initial segmentation, rather than relying on machine learning-based segmentation approaches, we adopt a physics-informed strategy that leverages the intrinsic coherence of the myocardium displacement field. Subsequently, these initial segmentations are used with a motion-guided segmentation algorithm and refined iteratively to generate robust segmentations for the full cardiac cycle. After extracting displacement information in the myocardial region, a computational model with denoising elements is applied to estimate a continuous displacement field from which cardiac strains are finally estimated.

The proposed method is first validated using an existing cylindrical phantom model, from which pseudo-experimental images are generated corresponding to the assigned phantom motion. These pseudo-experimental images are then used in the proposed pipeline and the computed displacements and strains are compared against the phantom ground-truth measures. The validation analyses are repeated for several signal to noise ratios (SNR), ranging from an ideal situation with no noise to more realistic scenarios with SNR values representative of regular clinical acquisitions. After successful validation, the proposed model is applied to preclinical images previously acquired in healthy swine subjects (all experiments were conducted under UCLA IACUC protocol ARC #2015-124). The results are finally compared with the ones computed using the DENSEanalysis toolbox (https://github.com/denseanalysis/denseanalysis), which is a common approach to process DENSE MRI data. Of particular interest in this comparison are the transmural variations of both displacements and strains: these may be key indicators of cardiac function, although they are often challenging to evaluate using traditional methods due to the limited number of voxels across the ventricular wall.

This pipeline is the necessary first step to link tissue displacements measurable via MRI to tissue function that depends on multiscale structures and mechanics.

Acknowledgements

This material is based upon work supported by the National Science Foundation under Award No. 2237391. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

Presenting Author: Uditha Weerasinghage University of Central Florida

Presenting Author Biography: Uditha Weerasinghage received his B.Sc. (Hons) degree in Civil Engineering from the University of Moratuwa, Sri Lanka, in 2022. He continued his studies at the same institution, where he finished his M.Sc. in Civil engineering. In 2023, he began his Ph.D. studies in Biomedical Engineering at the University of Central Florida in the Computational Biomechanics Lab. His research is in the areas of biomechanics and computational modeling, with a focus on cardiac function and cardiac microstructure.

Authors:

Uditha Weerasinghage University of Central Florida
Michael Freeland University of Central Florida
Luigi Perotti University of Central Florida