Quantitative Evaluation of the Morphological Heterogeneity in Breast Cancer Progression
Mark C Lloyd, Katarzyna Rejniak, Joseph O Johnson, Robert Gillies, Robert Gatenby, Marilyn M Bui. Moffitt Cancer Center, Tampa, FL
Background: Cancer cell heterogeneity is known as a factor of cancer progression and resistance to therapy. Quantitative evaluation of tumor heterogeneity have been carried out at the molecular and genetic scale; however, little is known at the cell morphology scale regarding to the variability of individual cells with respect to phenotypic core traits (proliferation, survival, morphology, and metabolism). We hypothesize that while genetics and signaling networks are the basis of core traits, cells' ability to perform core trait functions under diverse conditions within the physical microenvironment may decide their trends in tumor growth dynamics. This study is to investigate the feasibility of computational morphological analysis algorithms to predict the breast cancer progression.
Design: Multiple serial sections (4 µm) of 12 lobular and ductal breast carcinoma resection were collected retrospectively. Digitalized HE and IHC slides (10 biomarkers) were computationally investigate the morphological features of cancer cells on the whole digitalized slides. Sophisticated computer algorithms were developed to segment cancerous regions from the tumor's microenvironment and other non-malignant tissue structures. Every cancer cell was also segmented individually (∼1.5-3million/sample depending on the tumor size) and 5 morphological features were extracted from each cell (nuclear size, N:C, nuclear intensity, cyto intensity and Haralick texture). This expansive data set was parsed and interrogated using co-variant analyses to indicate if subpopulations of cells at the leading edge of invasive cancers were morphologically identifiable.
Results: A single morphological parameter was helpful in identifying subpopulaitons of cells spatially related to the invasive edge of GIII tumors, while not significantly present in GI samples. A multiparametric analysis of morphological features and molecular expression in the same sample indicated 2.6% (n=264) of the total cancer cell population within 250μm of the invasive edge of GI tumors, 14.7% (n=1,338) in GII, and 21.7% (n=19,584) in GIII. This represented GII>GI (5fold) and GIII>GII (15fold) in identifiable subpopulations of aggressive cells.
Conclusions: Morphological heterogeneity of breast cancer can be quantitatively evaluated by image analysis and are correlated with tumor progressiveness to aid in treatment descision making.
Tuesday, March 20, 2012 1:00 PM
Poster Session IV # 217, Tuesday Afternoon