FT-IR spectroscopic imaging is a powerful tool for the analysis of biological specimens as it reveals the vibrational structure of the tissue. High definition spectroscopic imaging further enables visualization at a level as small as a micron and label free extraction of biochemical information for efficient and accurate diagnosis. Here we couple high definition FT-IR spectroscopic imaging with the identification of biologically relevant spectral features and statistical analysis. The classification model is based on a learning algorithm that relates spectral features to established histologic classes. A random forest algorithm and a Bayesian algorithm are compared by the ROC values produced in their quantitative classification of normal, hyperplasia, dysplasia and malignant breast tissue, each scanned using a HD FT-IR spectroscopic imaging system with 1.1 µm spatial and 4 cm<sup>-1 </sup>spectral resolution. Classification results shown in fig 1 c) and f) demonstrate successful identification of epithelium, stroma, interlobular stroma, and myoepithelium which can be used as biomarkers for breast cancer detection and quantifying disease prognosis in terms of the relative distribution of these biomarkers in the tissue. The contrast of various cellular structures obtained with multiple immunohistochemical stains can be achieved in just one classified FT-IR spectroscopic image as shown in figure 1. HD breast tissue classification for automated histopathology will offer new opportunities for disease classification, diagnosis, and prognosis estimation.

Building a high definition breast tissue classifier: Applications in FT-IR histopathology

249th National Meeting (2015)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.