Using 3D Convex Hulls to Detect Coronary Artery Stenoses and Atherosclerotic Vessel Wall Lesions in Contrast Enhanced MDCT Images

A. Hans, K. Krissian, U. Joseph Schoepf, Ron Kikinis, C.-F. Westin
0002
April, 2004

Download full paper

Abstract

We present a new method for computing the convex hull of a structure. Based on a morphological definition of the convex hull as the intersection of all half-planes containing a structure, it can be computed by projecting all pixels to a line that is defining a certain half-plane. We use our method to detect stenoses and atherosclerotic plaques in grey scale cross-sections of vessels by comparing the convex hull to the original image and thus extracting the concavity regions. We apply it first to synthetic images to evaluate the general reliability of the algorithm, then to computed tomographic (CT) scans of the coronary arteries to evaluate its feasibility for real-life medical images.

Segments of the left (three left images) and right (two right images) coronary arteries of two patients. Possible plaques are detected and highlighted.

Reference

Hans A, Krissian K, Schoepf UJ, Kikinis R, Westin CF. Using 3d convex hulls to detect coronary artery stenoses and atherosclerotic vessel wall lesions in contrast enhanced mdct images. Technical Report 0002, Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, Laboratory of Mathematics in Imaging, 2004. ISSN.

Bibtex entry

@TechReport{hansLMI0002-04,
  author         = {Arne Hans and Karl Krissian and U. Joseph Schoepf and Ron  
                   Kikinis and Carl-Fredrik Westin},                           
  title          = {Using 3D Convex Hulls to Detect Coronary Artery Stenoses   
                   and Atherosclerotic Vessel Wall Lesions in Contrast Enhanced
                   MDCT Images},                                               
  institution    = {Department of Radiology, Brigham and Women's Hospital,     
                   Harvard Medical School},                                    
  year           = 2004,                                                       
  number         = 0002,                                                       
  address        = {Laboratory of Mathematics in Imaging},                     
  month          = {April},                                                    
  note           = {ISSN}
}

Grants

Siemens Medical Solutions

Research areas

Convex Hull, Segmentation, Vascular, CT