Classification of breast cancer grades through quantitative characterization of ductal structure morphology in three-dimensional cultures.

ABSTRACTA set of morphological features for mammary ductal structures is proposed for the evaluation of tissue's metastatic potential and exploited for the classification of six cell lines (MCF10 series) from three cancer grades in three-dimensional in vitro breast culture systems. These cultures mimic various grades of breast cancer along the metastatic cascade. Particular proteins of interest are fluorescently labeled and subsequently imaged via multi-spectral confocal microscopy in order to analyze cellular organization and polarity between various tumor grades. A Fourier-based image registration method is employed to stitch the images, 3D watershed segmentation method is used to identify the ductal structures, and morphological image processing techniques are utilized to extract the features. In addition to capturing both expected and visually-differentiable changes, we quantify subtle differences that are challenging to assess by microscopic inspection. Support vector machine-based supervised learning achieves 69.4% and 79.0% accuracy in classifying the ductal structures into pre-cancerous, non-invasive ductal carcinoma, and invasive ductal carcinoma grades after 7 and 14 days in the cell cultures, respectively. These results indicate that the proposed features can be used to describe the effects of morphological changes to the functional changes of the cells along the metastatic cascade.