Epithelial and Stromal Spectral Imaging for Rapid Surgical Margin Analysis

Abstract

A broadband spectroscopy platform was developed to image thick tissue samples at a resolution sensitive to the diagnostic gold standard, pathology. Tissue samples were imaged in a 1cm2 field of view across a static beam using a motorized stage, permitting wide-field optical characterization of diagnostic pathology. The sampling spot size confined the volume of tissue probed to within a few transport pathlengths so that multiple-scattering effects were minimized and simple empirical models parameterized the spectra. A k-Nearest Neighbor (k-NN) classifier was trained using parameters extracted from the localized scattering spectrum, automating diagnosis of benign and malignant breast pathologies in situ with a sensitivity and specificity of 91% and 77% respectively. Performance of the classifier was validated in 67,000 spectra from 29 excised breast tissues. The work effectively characterized the spectral response of breast pathologies and automated classification of the tissue's spectral response according to a diagnosis. Clinically feasible data acquisition speeds were attained through development of a dark-field in situ scanning-beam spectroscopy platform.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2011
Accession Number
ADA545295

Entities

People

  • Ashley M Laughney

Tags

DTIC Thesaurus Topics

  • Adipose Tissue
  • Breast Cancer
  • Broadband
  • Carcinoma
  • Classification
  • Histological Techniques
  • Histology
  • Institutional Review Board
  • Machine Learning
  • Mastectomy
  • Medical Personnel
  • Neoplasms
  • Pathology
  • Platforms
  • Scattering
  • Spectra
  • Spectroscopy

Fields of Study

  • Physics

Readers

  • Image Processing and Computer Vision.
  • Oncology (Cancer Research).
  • Spectroscopy.