Share Email Print

Proceedings Paper

Computerized analysis of sonograms for the detection of breast lesions
Author(s): Karen Drukker; Maryellen Lissak Giger; Karla Horsch; Carl J. Vyborny
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

With a renewed interest in using non-ionizing radiation for the screening of high risk women, there is a clear role for a computerized detection aid in ultrasound. Thus, we are developing a computerized detection method for the localization of lesions on breast ultrasound images. The computerized detection scheme utilizes two methods. Firstly, a radial gradient index analysis is used to distinguish potential lesions from normal parenchyma. Secondly, an image skewness analysis is performed to identify posterior acoustic shadowing. We analyzed 400 cases (757 images) consisting of complex cysts, solid benign lesions, and malignant lesions. The detection method yielded an overall sensitivity of 95% by image, and 99% by case at a false-positive rate of 0.94 per image. In 51% of all images, only the lesion itself was detected, while in 5% of the images only the shadowing was identified. For malignant lesions these numbers were 37% and 9%, respectively. In summary, we have developed a computer detection method for lesions on ultrasound images of the breast, which may ultimately aid in breast cancer screening.

Paper Details

Date Published: 9 May 2002
PDF: 5 pages
Proc. SPIE 4684, Medical Imaging 2002: Image Processing, (9 May 2002); doi: 10.1117/12.467093
Show Author Affiliations
Karen Drukker, Univ. of Chicago (United States)
Maryellen Lissak Giger, Univ. of Chicago (United States)
Karla Horsch, Univ. of Chicago (United States)
Carl J. Vyborny, Univ. of Chicago (United States)

Published in SPIE Proceedings Vol. 4684:
Medical Imaging 2002: Image Processing
Milan Sonka; J. Michael Fitzpatrick, Editor(s)

© SPIE. Terms of Use
Back to Top