Share Email Print

Proceedings Paper

Visualization and tissue classification of human breast cancer images using ultrahigh-resolution OCT (Conference Presentation)
Author(s): Xinwen Yao; Yu Gan; Ernest W. Chang; Hanina Hibshoosh; Sheldon Feldman; Christine P. Hendon

Paper Abstract

We employed a home-built ultrahigh resolution (UHR) OCT system at 800nm to image human breast cancer sample ex vivo. The system has an axial resolution of 2.72µm and a lateral resolution of 5.52µm with an extended imaging range of 1.78mm. Over 900 UHR OCT volumes were generated on specimens from 23 breast cancer cases. With better spatial resolution, detailed structures in the breast tissue were better defined. Different types of breast cancer as well as healthy breast tissue can be well delineated from the UHR OCT images. To quantitatively evaluate the advantages of UHR OCT imaging of breast cancer, features derived from OCT intensity images were used as inputs to a machine learning model, the relevance vector machine. A trained machine learning model was employed to evaluate the performance of tissue classification based on UHR OCT images for differentiating tissue types in the breast samples, including adipose tissue, healthy stroma and cancerous region. For adipose tissue, grid-based local features were extracted from OCT intensity data, including standard deviation, entropy, and homogeneity. We showed that it was possible to enhance the classification performance on distinguishing fat tissue from non-fat tissue by using the UHR images when compared with the results based on OCT images from a commercial 1300 nm OCT system. For invasive ductal carcinoma (IDC) and normal stroma differentiation, the classification was based on frame-based features that portray signal penetration depth and tissue reflectivity. The confusing matrix indicated a sensitivity of 97.5% and a sensitivity of 77.8%.

Paper Details

Date Published: 19 April 2017
PDF: 1 pages
Proc. SPIE 10053, Optical Coherence Tomography and Coherence Domain Optical Methods in Biomedicine XXI, 100532A (19 April 2017); doi: 10.1117/12.2254842
Show Author Affiliations
Xinwen Yao, Columbia Univ. (United States)
Yu Gan, Columbia Univ. (United States)
Ernest W. Chang, Columbia Univ. (United States)
Hanina Hibshoosh, Columbia Univ. (United States)
Sheldon Feldman, Columbia Univ. (United States)
Christine P. Hendon, Columbia Univ. (United States)

Published in SPIE Proceedings Vol. 10053:
Optical Coherence Tomography and Coherence Domain Optical Methods in Biomedicine XXI
James G. Fujimoto; Joseph A. Izatt; Valery V. Tuchin, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?