Computer Prognostic Model Predicts Breast Cancer Survival

Incorporation of features from the breast cancer epithelium and stroma improves prognosis

THURSDAY, Nov. 10 (HealthDay News) -- A computational system-based prognostic model incorporating breast cancer stromal and epithelial features can be used as a strong predictor of breast cancer survival, with stromal features more predictive of survival than epithelial ones, according to a study published in the Nov. 9 issue of Science Translational Medicine.

Andrew H. Beck, M.D., from Stanford University School of Medicine in California, and colleagues sought to develop a high-accuracy, image-based predictor of survival in breast cancer by analyzing new clinically predictive morphologic phenotypes of breast cancer. A computational pathologist (C-Path) system was developed to identify a quantitative feature set of 6,642 breast cancer epithelium and stromal features, including both standard morphometric image object descriptors and higher-level contextual, relational, and global image features. These measurements were used to develop a prognostic model score, which was applied to C-Path microscopic images from 248 breast cancer patients from the Netherlands Cancer Institute (NKI), and 328 patients from the Vancouver General Hospital (VGH).

The investigators found a strong and significant correlation between the prognostic model score generated by the C-Path system and overall survival in both the NKI and the VGH cohorts, which was independent of clinical, pathological, and molecular factors. A significant correlation was found between three stromal features and survival, and this correlation was stronger than the association between epithelial characteristics in the model and survival.

"These findings implicate stromal morphologic structure as a previously unrecognized prognostic determinant for breast cancer," the authors write.

Full Text (subscription or payment may be required)
Editorial (subscription or payment may be required)

Copyright © 2011 HealthDay. All rights reserved.

Powered by