Files in this item



application/pdfECE499-Sp2016-moscoso.pdf (2MB)Restricted to U of Illinois
(no description provided)PDF


Title:Cell classification for imaging-genomics analysis of breast cancer
Author(s):Moscoso, Miguel
Contributor(s):Do, Minh N.
Degree:B.S. (bachelor's)
support vector machines
computational pathology
Abstract:Imaging-genomics aims to combine two separate medical modalities, imaging and genomics, to obtain more accurate and insightful prognoses. We use breast cancer histopathologic images and their features to determine how they are an expression of the underlying genotype. Our approach is to build an image analysis pipeline that consists of five key steps. First, we segment the image into nuclei and their corresponding cells, from which we extract features describing the cells shape, color, and texture. We then use these features to train a support vector machine classifier to allow for proper labeling of cells, specifically epithelial and stroma cells. The classification of cells allows us to understand spatial layout and morphometry, which are important prognosis indicators. Next, we collect high-level image features, to correlate with genomic data for new understandings and computational prognosis.
Issue Date:2016-05
Genre:Dissertation / Thesis
Date Available in IDEALS:2016-08-30

This item appears in the following Collection(s)

Item Statistics