Title:
Mesh extraction from 4D transesophageal echocardiogram for patient-specific digital simulation
Mesh extraction from 4D transesophageal echocardiogram for patient-specific digital simulation
Author(s)
Gunther, Matthew Joseph
Advisor(s)
Dasi, Lakshmi
Anderson, David V.
Anderson, David V.
Editor(s)
Collections
Supplementary to
Permanent Link
Abstract
Mitral valve (MV) regurgitation is a condition resulting from structural, positional,
and/or movement abnormalities of the MV. This valvular heart disease (VHD) is characterized by a regurgitant jet into the left atrium (LA) upon contraction of the left ventricle
(LV). Because many pathologies may cause the same symptoms of incomplete closure
of the MV leaflets, patient-specific diagnostic tools can greatly aid treatment planning.
Three-dimensional (3D) transesophageal echocardiogram (TEE), one of the primary diagnostic tools for MV degeneration, can be leveraged beyond an observational imaging
platform for patient-specific digital simulation of the MV anatomy. This thesis proposes a
computational pipeline aimed at automating the deformation of a single MV mesh across
TEE frames of the cardiac cycle. Linear filtering techniques are used to reduce noise and
segment the anatomy within each volumetric frame to create a mesh representation of the
cardiac anatomy. Two algorithms are compared for mesh deformation across frames in
a TEE series: one that employs active surface modeling exclusively, and another which
transitions from dense optical flow to active surface modeling. The comparative analysis
of these algorithms was conducted using TEE series from 10 different patients. The results indicate a higher efficacy of the composite algorithm. By first deforming the mesh
using the general motion of the structures in the image and then snapping the faces of the
deforming mesh to the target volume, more accurate deformations were achieved. One of
the primary limitations of the algorithm is the dependency on the quality and resolution of
3D TEE data, which can vary significantly across captures. Future improvements are suggested to optimize the automated mesh extraction and preserve the mesh structure during
deformation.
Sponsor
Date Issued
2023-12-08
Extent
Resource Type
Text
Resource Subtype
Thesis