Takeshi Hara PhD, Presenter: Nothing to Disclose

Nanae Yagi, Abstract Co-Author: Nothing to Disclose

Tomoko Matsubara PhD, Abstract Co-Author: Nothing to Disclose

Hiroshi Fujita PhD, Abstract Co-Author: Nothing to Disclose

Tokiko Endo MD, Abstract Co-Author: Nothing to Disclose

Masatoshi Tsuzaka PhD, Abstract Co-Author: Nothing to Disclose

Takuji Iwase MD, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

Architectural distortion is a common cause of false-negative findings on screening mammograms because of its subtlety and potential for malignancy. We have developed the detection algorithm for distorted area with focal retraction around skinline in order to aid physicians with diagnosis. However, previous CAD studies have focused either on microcalcifications or breast masses or both. The purpose of our study is to suggest the automated detection method for another architectural distortion within mammary gland.

The distributions of mammary gland are approximated to linear structures. Those directions are toward the nipple within normal breast, whereas those ones are toward the suspect area within abnormal breast. First, the linear structures of mammary gland are extracted by the mean curvature. The suspect area is then determined by concentration index. This index is obtained from the circle and half-circle regions in order to detect the distorted areas both within and around border of mammary gland. Finally, the false positives are reduced by the discriminant analysis by using size, mean pixel value, mean concentration index, mean isotropy index, contrast between the suspected region and surrounding region, root mean square variation, the first moment, and two features that indicated symmetry in the power spectrum.Our image database of architectural distortions consists of 94 cases. Forty-one cases contain concentrations of mammary structures within mammary glandular tissue.

Our image database of architectural distortions consists of 94 cases. Forty-one cases contain concentrations of mammary structures within mammary glandular tissue. As a result, the performance of our method indicated a sensitivity of 80% (33/41) with 0.9 false positive per image.

The results of our preliminary study showed that our approach may be effective, since it detected architectural distortions accurately and at a moderate sensitivity while producing few false positives per image.

Hara, T, Yagi, N, Matsubara, T, Fujita, H, Endo, T, Tsuzaka, M, Iwase, T, et al, , Automated Method for Detecting Mammographic Architectural Distortion Based on Concentration of Mammary Gland.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4407920.html