Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Semi-automatic spatio-temporal reconstruction of the left ventricle from CMR
Cardiovascular simulation and modelling have recently grown into a promising area of research, with a major interest in simulating biophysical phenomena within patient specific environments [1, 2, 3, 4]. With a major focus being faster, less computationally expensive (approaching real time) methods...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
AUC Knowledge Fountain
2019
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| Summary: | Cardiovascular simulation and modelling have recently grown into a promising area of research, with a major interest in simulating biophysical phenomena within patient specific environments [1, 2, 3, 4]. With a major focus being faster, less computationally expensive (approaching real time) methods along with reasonable accuracy to actually translate research to application in clinical decision making. This work presents a quick computational framework for building a semi-automatic platform building patient specific spatio-temporal kinematic models of the left ventricle, allowing interactive visualization and detailed quantification of structural e.g. (volume, wall thickness) and functional parameters e.g. (radial Strain, circumferential Strain) of the beating heart. The proposed method relies on initial segmentation and two-dimensional registration of the region of interest from a set Magnetic Resonance Steady State Free Procession (MRSSFP) images in different orientations. Then, a method herein is developed to reconstruct a three-dimensional volumetric, tetrahedral mesh using segmentation information, and a spatial interpolation scheme is used that interpolates the 2D displacement fields to deform the initial 3D volume mesh over the cardiac cycle. The method was tested using the Statistical Atlases and Computational Models of the Heart (STACOM 2011) motion tracking challenge dataset [5]. After analysis, the chamber volumes, peak strains and tracking errors were compared against the results of other participants in the challenge and reference values from literature. We report an average median tracking error for all datasets of (4. 9 ± 0. 82mm), end-diastolic volume of (123 ± 21ml), end-systolic volume of (66 ± 13ml). Average peak radial Strain of (31 ± 9%), circumfriential strain of (−14 ± 1. 2%) and longitudinal Strain of (−12 ± 7. 3%), with running time per case, approximately 15 minutes, including the manual pre-processing steps. The findings indicate that although the proposed method allows the efficient estimation of physiollogically plausible volume and strain curves, with tracking error well within reported findings using the same dataset and image modality, reliance on SSFP images alone yields underestimation of circumfriential and longitudinal deformation components which may indicate the need of additionally incorporating displacement information from Tagging MRI in our interpolation scheme which complements the ability of SSFP images in tracking deformation. The speed and accuracy provided by our computational platform facilitates it’s use for clinical assistance in diagnosis, pre-operative surgery planning and post-operative assesment. And through the reliance on streamlined data flow and efficent, scalable VTK format for data export, it can be further used to complement and improve other computational and statistical models of the dynamic heart. |
|---|