Design
Embedded Files
Problem Statement
Most of the research done on atherosclerosis etiology has focused solely on the lesions themselves or the inside of blood vessels. None of this research has led to the discovery of the full etiology of this disease. This has caused a large push for research on the tissues that surround the lesions on the outside of the blood vessel to hopefully provide new insight on the pathophysiology of atherosclerosis. The majority of this analysis relies on CT images to view the outside of the blood vessels but there is currently no effective image processing tool that is user friendly and accurate enough to utilize.
Design Objectives
Improve the current MATLAB algorithm for data processing
Increase current knowledge of surrounding tissue
Easy usage and interpretation of the algorithm's results
Perform quantitative analysis to increase knowledge of the adventitial tissue
Understand if epicardial adipose tissue (EAT) is related to lesion formation in the coronary artery
Overall Design Solution
Phase 1: Lesion Identification and Segmentation
Phase 1: Lesion Identification and Segmentation
Manually identified lesion sites in CT images. Identified at dark portions of the vessel using HOROS.
Used ITK-SNAP to segmnet the region of interest to isolate the area of interest.
Phase 2: Uploading Images to MATLAB
Phase 2: Uploading Images to MATLAB
This step loads DICOM and NIfTI images onto MATLAB.
The images were stored as 3D matrices. DICOM images were made up of values that corresponded to the pixel intensity levels in Hounsfield units, while NIfTI images had the number '1' as an entry for segmented regions and '0' for non-segmented regions.
NIfTI images also had their coordinate scheme re-adjusted to match that of the DICOM images.
Phase 3: Tissue Identification
Phase 3: Tissue Identification
- Used the matrices from Phase 2 in order to identify voxel value correspondence to tissue type.
- Simplified the matrix by relabeling the Hounsfield values into the corresponding tissue types.
Phase 4: Centerline Extraction, Linear Approximation, & Color Mapping
Phase 4: Centerline Extraction, Linear Approximation, & Color Mapping
Centerline Extraction
Determine the center point of each slice in the 3D image matrix
Output a visualization of the segmented vessel
Major drawback:
Cannot work at bifurcations
Skeleton 3D program used was originally designed for bone
Linear Approximation
Determine principal direction of movement at each center point by finding a vector from said center point that reaches the next 2 center points.
Fit two separate polynomials as parametric curves in the non-principal directions determined in the prior step.
Use the derivatives of the polynomials to determine a normal vector that's perpendicular to the direction of movement, then save 2D slice of the vessel at that center point as a plane perpendicular to the normal vector.
After completing this for each point, stack perpendicular planes to reveal a linearized vessel.
Characterization
Using the same HU thresholds from the Tissue Idenitification function, generate a custom color map to be used in MATLAB's Volume Viewer.
After loading vessel into Volume Viewer, select the "Colormap" option and change the color scheme to "workspace variable".
Phase 5: Tissue Distribution
Phase 5: Tissue Distribution
Using TID.m from Phase 3, each tissue type percentage was found.
Found the percentage of tissues surrounding the vessel: muscle, adventitia, fat, calcium.
Calculated from the number of voxels each tissue spanned, respectively, divided by the total number of voxels in our region of interest.
UC San Diego 2021
Page Leaders: Vanessa Rohrer / Danielle Manalastas
Page updated
Report abuse