Share Email Print
cover

Proceedings Paper

Automated adipose study for assessing cancerous human breast tissue using optical coherence tomography (Conference Presentation)
Author(s): Yu Gan; Xinwen Yao; Ernest W. Chang; Syed A. Bin Amir; Hanina Hibshoosh; Sheldon Feldman; Christine P. Hendon
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Breast cancer is the third leading cause of death in women in the United States. In human breast tissue, adipose cells are infiltrated or replaced by cancer cells during the development of breast tumor. Therefore, an adipose map can be an indicator of identifying cancerous region. We developed an automated classification method to generate adipose map within human breast. To facilitate the automated classification, we first mask the B-scans from OCT volumes by comparing the signal noise ratio with a threshold. Then, the image was divided into multiple blocks with a size of 30 pixels by 30 pixels. In each block, we extracted texture features such as local standard deviation, entropy, homogeneity, and coarseness. The features of each block were input to a probabilistic model, relevance vector machine (RVM), which was trained prior to the experiment, to classify tissue types. For each block within the B-scan, RVM identified the region with adipose tissue. We calculated the adipose ratio as the number of blocks identified as adipose over the total number of blocks within the B-scan. We obtained OCT images from patients (n = 19) in Columbia medical center. We automatically generated the adipose maps from 24 B-scans including normal samples (n = 16) and cancerous samples (n = 8). We found the adipose regions show an isolated pattern that in cancerous tissue while a clustered pattern in normal tissue. Moreover, the adipose ratio (52.30 ± 29.42%) in normal tissue was higher than the that in cancerous tissue (12.41 ± 10.07%).

Paper Details

Date Published: 19 April 2017
PDF: 1 pages
Proc. SPIE 10043, Diagnosis and Treatment of Diseases in the Breast and Reproductive System, 100430G (19 April 2017); doi: 10.1117/12.2253263
Show Author Affiliations
Yu Gan, Columbia Univ. (United States)
Xinwen Yao, Columbia Univ. (United States)
Ernest W. Chang, Columbia Univ. Medical Ctr. (United States)
Syed A. Bin Amir, Columbia Univ. (United States)
Hanina Hibshoosh, Columbia Univ. Medical Ctr. (United States)
Sheldon Feldman, Columbia Univ. Medical Ctr. (United States)
Christine P. Hendon, Columbia Univ. (United States)


Published in SPIE Proceedings Vol. 10043:
Diagnosis and Treatment of Diseases in the Breast and Reproductive System
Melissa C. Skala; Paul J. Campagnola, Editor(s)

© SPIE. Terms of Use
Back to Top