Date on Master's Thesis/Doctoral Dissertation

8-2012

Document Type

Master's Thesis

Degree Name

M. Eng.

Department

Electrical and Computer Engineering

Committee Chair

Farag, Aly A.

Author's Keywords

Virtual colonoscopy; Colon segmentation; Colonoscopy; Medical visualization; Colonscopy CAD system; Colon centerline

Subject

Colon (Anatomy)--Tomography; Diagnostic imaging--Digital techniques

Abstract

Colorectal cancer is the fourth most common cancer, and the fourth leading cause of cancer related death in the United States. It also happens to be one of the most preventable cancers provided an individual performs a regular screening. For years colonoscopy via colonoscope was the only method for colorectal cancer screening. In the past decade, colonography or virtual colonoscopy (VC) has become an alternative (or supplement) to the traditional colonoscopy. VC has become a much researched topic since its introduction in the mid-nineties. Various visualization methods have been introduced including: traditional flythrough, colon flattening, and unfolded-cube projection. In recent years, the CVIP Lab has introduced a patented visualization method for VC called flyover. This novel visualization method provides complete visualization of the large intestine without significant modification to the rendered three-dimensional model. In this thesis, a CVIP Lab VC interface was developed using Lab software to segment, extract the centerline, split (for flyover), and visualize the large intestine. This system includes adaptive level sets software to perform large intestine segmentation, and CVIP Lab patented curve skeletons software to extract the large intestine centerline. This software suite has not been combined in this manner before so the system stands as a unique contribution to the CVIP Lab colon project. The system is also a novel VC pipeline when compared to other academic and commercial VC methods. The complete system is capable of segmenting, finding the centerline, splitting, and visualizing a large intestine with a limited number of slices (~350 slices) for VC in approximately four and a half minutes. Complete CT scans were also validated with the centerline extraction external to the system (since the curve skeletons code used for centerline extraction cause memory exceptions because of high memory utilization).

Share

COinS