Meeting Banner
Abstract #4935

Automatic Breast Tumor Segmentation Methods for Mass and Non-mass Lesions for Quantitative Morphology and Texture Analysis

Xinxin Wang1, Yang Zhang1, Jeon-Hor Chen1,2, Siwa Chan3, and Min-Ying Su1

1University of California, Irvine, Irvine, CA, United States, 2E-Da Hospital and I-Shou University, Kaohsiung, Taiwan, 3Tzu-Chi General Hospital, Taichung, Taiwan

A breast tumor segmentation platform for mass and non-mass tumors on 3D MRI was developed. The segmentation of non-mass lesions is challenging. We developed a new method based on region-growing with the threshold determined by comparison of the intensity histograms in an ROI containing suspicious tumor region vs. outside ROI containing normal fibroglandular tissues. Breast MRI of 122 patients with pathologically-confirmed breast cancer were studied. Of them, 14 had triple negative, 29 had HER2-positive, and 51 had Hormonal-positive, HER2-negative breast cancers. The segmented tumor ROI was analyzed to obtain morphology and texture parameters for differentiation of these 3 molecular subtypes.

This abstract and the presentation materials are available to members only; a login is required.

Join Here