Abstract #2885

Chen W, Giger M

We extend the conventional 2D texture analysis technique to 3D using gray-level co-occurrence matrix (GLCM) method, and assess the performance of textural features in the task of distinguishing between malignant and benign breast lesions in DCE-MRI. There are two advantages in our technique: First, the counting statistics are improved in 3D analysis. Second, our analyses were performed on segmented 3D lesions, which render feature extraction not affected by normal tissues surrounding the lesions. We found 3D texture analysis based on accurately segmented 3D breast lesions improved diagnostic accuracy as compared to 2D texture analyses based on 2D ROIs.