Label-Free High-Resolution Infrared Spectroscopic Imaging for Breast Tissue Histopathology
Michael J Walsh, Sreeradha Biswas, Andre Kajdacsy-Balla, Rohit Bhargava. University of Illinois at Urbana-Champaign, Urbana, IL; University of Illinois at Chicago, Chicago, IL
Background: Histopathology forms for the gold standard for breast cancer diagnosis. Current methodologies require taking multiple tissue sections from a tissue block which are subsequently stained using special and immunohistochemical (IHC) stains to identify cell types and disease state. Infrared (IR) spectroscopic imaging is a novel approach to deriving cell type and disease status information in an entirely label-free non-perturbing approach from a single unstained tissue section based on the inherent biochemistry of the tissue.
Design: We have built the first ever high-resolution IR imaging instrument capable of imaging in a clinical setting at a spatial resolution of 1µm x 1µm. Previous results demonstrated accurate histological segmentation of breast tissue however this was limited to large cell types. We have imaged at a high-resolution a training array consisting of 50 patient biopsies and a validation array of 50 patient biopsies, with tissues including normal, hyperplasia, dysplasia and cancer. A Bayesian classifier was built to accurately segment the IR data into over ten epithelial and stromal cell types from a single unstained biopsy.
Results: Accurate histological segmentation from an IR image was performed for epithelial and stromal cell types, including epithelial cells, fibroblasts, cancer-activated fibroblasts, lymphocytes, collagen and necrosis. This study also demonstrated for the first time the identification, chemical characterization, and classification of myoepithelial cells and endothelial cells in breast tissue biopsies using IR imaging. IR spectroscopic imaging coupled with Bayesian classifiers demonstrated a high degree of cell typing accuracy with an Area Under the ROC Curve (AUC) of over 0.9.
Conclusions: IR imaging coupled with spectral classifiers represents a novel approach to derive cell type and disease status information from tissue based on tissue biochemistry without the requirement for staining. IR imaging has the potential to be a powerful adjunct to the current clinical workflow, with the ability to give additional information about cell types present without requiring additional analysis. In addition, IR imaging has the potential to give insight into chemical changes associated with breast cancer progression.
Tuesday, March 5, 2013 1:15 PM
Proffered Papers: Section E, Tuesday Afternoon