Share Email Print
cover

Proceedings Paper

Automatic location of cylindrical bones in digital projection radiographs using eigenvector analysis
Author(s): Susan S. Young; Hsien-Che Lee
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Image processing algorithms that automatically locate patterns in digital images have application to medical image enhancement processing and computer-aided diagnosis (CAD). A flexible and computationally efficient detection and classification algorithm has been developed that can be readily adapted for any specified anatomical structure or imaging modality. The algorithm makes use of image pattern filters and eigenvector analysis. The algorithm was optimized and tested for the detection of the class of cylindrical bones in digital projection radiography. Gradient-based edge detection is used to locate candidate bone edge pixels. For each candidate pixel, the bone image profile (i.e., the code values on a line segment that cross the bone perpendicular to the bone edge) is analyzed to identify indicators of bone location. Shape constraints of the image profile are used to refine the selection of candidate profiles. An eigenvector representation for the cylindrical bone profiles is then constructed. The algorithm then automatically classifies structures in the imagery that matches the eigenvector representation of the cylindrical bone profile. Using an eigenvector representation of a cylindrical bone profile constructed from a set of training images, the algorithm correctly classified 91 humerus bone profiles that were tested.

Paper Details

Date Published: 6 June 2000
PDF: 12 pages
Proc. SPIE 3979, Medical Imaging 2000: Image Processing, (6 June 2000); doi: 10.1117/12.387672
Show Author Affiliations
Susan S. Young, Eastman Kodak Co. (United States)
Hsien-Che Lee, Eastman Kodak Co. (United States)


Published in SPIE Proceedings Vol. 3979:
Medical Imaging 2000: Image Processing
Kenneth M. Hanson, Editor(s)

© SPIE. Terms of Use
Back to Top