Optical Biopsy Method for Breast Cancer Diagnosis Based on Artificial Neural Network Classification of Fluorescence Landscape Data

Abstract

Supervised self-organizing map, a type of artificial neural network, is applied for classification of human breast tissue samples utilizing data obtained from fluorescence landscape measurements. Female breast tissue samples were taken soon after the surgical resection, identified and stored at -80 degrees C until fluorescence measurements. From fluorescence landscapes obtained in UV-VIS region spectral features showing statistically significant differences between malignant and normal samples are identified and further quantified to serve as a training input to neural network. Additional set of samples was used as a test group input to trained network in order to evaluate performance of proposed optical biopsy method. Classification sensitivity of 83.9% and specificity of 88.9% are found.