COMPUTERIZED ANALYSIS OF BREAST-LESIONS IN 3 DIMENSIONS USING DYNAMICMAGNETIC-RESONANCE-IMAGING

Contrast-enhanced magnetic resonance imaging (MRT) of the breast is kn own to reveal breast cancer with higher sensitivity than mammography a lone. The specificity is, however, compromised by the observation that several benign masses take up contrast agent in addition to malignant lesions. The aim of this study is to increase the objectivity of brea st cancer diagnosis in contrast-enhanced MRI by developing automated m ethods for computer-aided diagnosis. Our database consists of 27 MR st udies from 27 patients. In each study, at least four MR series of bath breasts are obtained using FLASH three-dimensional (3D) acquisition a t 90 s time intervals after injection of Gadopentetate dimeglumine (Gd -DTPA) contrast agent. Each series consists of 64 coronal slices with a typical thickness of 2 mm, and a pixel size of 1.25 mm. The study co ntains 13 benign and 15 malignant lesions from which features are auto matically extracted in 3D. These features include margin descriptors a nd radial gradient analysis as a function of time and space. Stepwise multiple regression is employed to obtain an effective subset of combi ned features. A final estimate of likelihood of malignancy is determin ed by linear discriminant analysis, and the performance of classificat ion by round-robin testing and receiver operating characteristics (ROC ) analysis. To assess the efficacy of 3D analysis, the study is repeat ed in two-dimensions (2D) using a representative slice through the mid dle of the lesion. In 2D and in 3D, radial gradient analysis and analy sis of margin sharpness were found to be an effective combination to d istinguish between benign and malignant masses (resulting area under t he ROC curve: 0.96). Feature analysis in 3D was found to result in hig her performance of lesion characterization than 2D feature analysis fo r the majority of single and combined features. In conclusion, automat ed feature extraction and classification has the potential to compleme nt the interpretation of radiologists in an objective, consistent, and accurate way. (C) 1998 American Association of Physicists in Medicine .