Level set breast mass segmentation in contrast-enhanced and non-contrast-enhanced breast CT

Abstract

Dedicated breast CT (bCT) is an emerging technology that produces 3D reconstructed images of the breast, thus allowing radiologists to detect and evaluate breast lesions in 3D. In previous work, we have developed an algorithm that combines radial gradient index (RGI) segmentation and a modified level set model for segmentation of lesions in contrast-enhanced bCT images; yielding an average overlap ratio (OR) of 0.69, which is higher than 0.4, the overlap ratio that is generally deemed "acceptable". In this study, this segmentation algorithm, with the same parameter settings, was applied to the corresponding non-contrastenhanced bCT images. The results show that the OR obtained on non-contrast images was 0.62, with the segmented lesion volumes tending to be slightly smaller as compared with those obtained on the corresponding contrast-enhanced images. These results imply that while use of contrast improves segmentation performance, the increase may not be significant, and thus, the role of noncontrast-enhanced breast CT should be further investigated. © 2012 Springer-Verlag Berlin Heidelberg.