|About GSBS | FAQ | Job Opportunities | Search UMDNJ|
Shanghai JiaoTong University
People`s Republic of China
Graduate Program in Biomedical Engineering
Thesis Advisor: Peter Yim, PhD
G1 Conference Room
Robert Wood Johnson Medical School
1 Robert Wood Johnson Place
New Brunswick, NJ 08903
Wednesday, April 29, 2009
The 3D segmentation technology is a useful facility helping radiologists to diagnose and analyze diseases including tumors. Thus it has vast potential value for radiologists.
The doctoral project is aiming at designing a model for reconstructing surface and segmenting medical objects such as livers and vessels from Computed Tomography (CT).
There are two main objectives. The first objective is relative to surface reconstruction utilizing orthogonal contours to realize a fast semi-automatic 3D surface reconstruction method. The other objective is relative to automatic 3D segmentation of livers in CT based on the prior knowledge from a training data set. The prior knowledge includes a diversity of features containing voxel intensity, position, texture and shape.
The semi-automatic algorithm for surface reconstruction from orthogonal contours contains three necessary steps: 1) detection of intersections between orthogonal contours which do not in practice perfectly intersect, 2) assembling contour segments into a coherent 3D polygonal mesh of the surface, 3) triangulation of 3D polygons with dynamic programming. The algorithm has produced good results on vessels and livers.
The automatic algorithm for livers segmentation contains three steps: 1) pre-processing and initial position detection, 2) optimized boundary classification 3) shape modeling based on robust 3D points registration and deformation gradients. Some experiments show the hybrid method can handle large variation exhibited in the shape distribution of livers and intensity similarity of adjacent organs.