10508 Background: Advances in high-resolution, real-time, optical imaging have enabled optical coherence tomography (OCT) for non-excisional optical biopsies of breast tissue. OCT is the optical analogue to ultrasound imaging, with resolution approaching that of histology. In breast tissue, regions of tumor, tumor margins, abnormal ducts, and foci of tumor cells can be identified based on increased scattering and morphological appearance. We have developed a portable clinical OCT system along with image criteria for identifying breast cancer in real-time during needle-biopsy and surgical-biopsy procedures. Methods: Over 4,000 depth-resolved OCT axial scans were extracted from over 50 cross-sectional OCT images from 9 ex vivo specimens of ductal carcinoma in situ (DCIS) of the human breast. OCT image data was acquired in real-time within 6 hours of surgical resection using a state-of-the-art OCT system with micron resolution. Imaged sites were marked for registration with hematoxylin and eosin-stained histology. Axial scan data was analyzed using frequency-domain analysis techniques to classify regions as tumor, stroma, or adipose tissue. OCT images of tissue types were compared with histological observations. Results: Real-time OCT images of microscopic breast cancer morphology showed strong correlations with corresponding histology, including regions of tumor, tumor margins, ducts, and adipose tissue. Computational frequency-domain analysis of depth-resolved axial scan data classified tissue types as tumor, stroma, and adipose with 91% sensitivity and 87% specificity. Conclusions: Optical biopsy of breast cancer with OCT is feasible for imaging at resolutions approaching histology. Frequency- domain analysis of axial scan data can be used to classify breast tissue types with a sensitivity and specificity that are comparable to or surpasses that of conventional mammography and ultrasound imaging of human DCIS. Clinical OCT imaging has the potential for real-time guidance and diagnostics in needle- or surgical-biopsy of breast cancer. No significant financial relationships to disclose.
Nanoparticle Classification Using Frequency Domain Analysis on Resource-Limited Platforms
