blood vessel modeling using 2d/3d level set method

Blood Vessel Modeling using2D/3D Level Set Method

Overview

Main Idea Big Picture Problem Considered Motivation Main Approach : Level-Set Based 2D Segmentation

Discussion Strength and Weakness of the Approach Comparison with other approaches Future Direction

Pipeline for Blood Flow Modeling

3D Image Acquisition

Geometric Modeling

Finite ElementCalculation

Quantification,Visualization

cardivascular disease research, medical device design, and surgical planning

What is Problem?

Construct accurate patient-specific 3D geometric models (tetrahedral mesh) of blood vessels of interest

Why is it important?

Flow calculation using finite element method requires quality mesh

Accuracy of calculation depends on that of geometric model

Geometry Quantification

from Image to GeomtryWhy is it difficult ? Usually requires ill-defined segmentation process

Computers easily make an error Thresholding (Isocontouring) is well defined but often it is not useful in

medical applications Noisy Scanning

Accurate Construction

Patient-Specific

Operator Dependence Minimal user interaction is favorable

Robustness Always guarantee quality result ?

Main Approach for Geometry Path (Centerline) Construction

Semi-automatic , manual

2D image Segmentation

Surface Reconstruction thru lofting

Solid Model Operation (CSG) : Union

Automatic Mesh Generation

Level Set Method for 2D Segmentation

Implicit Geometry Description (X) = 0 ex) x,y,z) = x + y + z -1 = 0

Evolving Geometry : (X,t)=0 t + v | grad = 0 Intuitively, move a lot on low intensity gradient area and move little on

high intensity gradient area along normal direction v : speed function , k : curvature , I : intensity

Segmentation Quality Measurement Compare segmentation with phantom geometry

2 2 2

Surface Reconstruction Lofting

Multiple Vessels - Boolean Set Operation (Union)

Automatic Mesh Generation from solid models

3D Level Set Method

Extend 2D to 3D No centerline necessary No lofting (interpolation) : more accurate Smooth Junction Often Hard to control (geometry & topology)

2D level set approach

Strength Intuitive : taking

advantage of tubular shape of vessels

Clean Output Controllability

Geometry Topology

Noise Removal Robust

Weakness Heavy manual work

Centerline construction A bunch of 2D segmentation

which is not automatic Stopping criteria

Inaccurate Lofting & interpolation

Non-smooth junction Possible Self-Intersection Limited to only tubular

structure segmentation

Comparison

2D

Segmentation

3D

Segmentation

Control Good Bad

Noise Tolerance Good Bad

Robustness Good Bad

Accuracy Bad Good

User Dependence Bad Good

Smoothness Good/Bad Good/Bad

Conclusion

Image Based Vascular Model Construction Centerline construction Level set based 2D segmentation along centerline Surface reconstruction thru lofting Solid modeling & mesh generation

Tradeoffs : 2D vs 3D segmentation