Quantitative differentiation of breast lesions based on three-dimensional morphology from magnetic resonance imaging

Objective: The purpose of this study was to develop geometric indices, based on morphologic measurements, of contrast-enhanced breast lesions using threedimensional magnetic resonance imaging (MRI) that can statistically differentiate between benign and malignant lesions. Methods: Thirty-three patients with abnormal mammograms were scanned using three-dimensional MRI. The lesion was then extracted from the volume data set using interactive volume-rendering software. From the extracted region of interest, the surface boundary corresponding to the lesion was extracted using an isosurface method. The boundary was represented as a triangular mesh, from which the surface area and volume enclosed by the surface were computed. The following variables were tested for discrimination ability between benign and malignant lesions: 1) the volume-tosurface area (V/S) ratio, 2) spherical shape index (SSI), and 3) resolution reduction for V/S and SSI values (100%, 16%, 8%, 4%, 2%, and 1% resolution). In addition, comparisons were made between the V/S and SSI values at 100% versus their reduced resolutions. Results: Thirty-three subjects (15 malignant tumors and 18 benign tumors) were studied. The SSI index (P = 0.0063) was a significant discriminator for malignancy. The V/S ratio (P = 0.9280) did not seem to be a useful variable in distinguishing benign and malignant masses. The V/S resolution comparisons (P > 0.5897) and the SSI resolution comparisons (P > 0.05) were also not useful discriminating variables. Conclusion: The SSI seems to be a useful factor in differentiating between benign and malignant lesions. Current clinical interpretation of breast lesions using MRI may be enhanced by the adjunctive use of this objective postanalysis method.