CRImage

CRImage analyzes H&E-stained tissue-section images to classify cells, quantify tumor cellularity, characterize spatial patterns, and integrate image-derived features with gene expression to improve interpretation of tumor heterogeneity and predict survival in estrogen receptor–negative breast cancer.

Key Features:

• Cell Classification: Processes H&E-stained images to classify distinct cell types within tissue sections for discrimination of cancerous and normal cells.
• Tumor Cellularity Calculation: Computes the proportion of cancer cells in a sample to deconvolute cellular heterogeneity and improve comparability of copy number profiles across samples.
• Integration with Molecular Data: Integrates image-derived features with gene expression data to build predictive models, including a predictor for survival in estrogen receptor–negative breast cancer that outperforms classifiers based solely on microarray expression signatures.
• Spatial Pattern Analysis: Quantitatively analyzes spatial organization and interactions, including stromal cell relationships, to reveal prognostic features not captured by molecular assays alone.

Scientific Applications:

• Breast cancer prognostic modeling: Applied to cohorts of breast tumors (323 discovery, 241 validation) to develop and validate image-informed survival predictors.
• Refinement of molecular profile interpretation: Uses image-based cellularity and spatial metrics to refine interpretation of gene expression and copy number data in heterogeneous tumor samples.
• Detection of subtle genomic aberrations: Enhances sensitivity for detecting genomic aberrations by accounting for tumor cellularity and cellular composition derived from H&E images.

Methodology:

Algorithmic analysis of standard H&E-stained tissue sections to extract quantitative metrics of cellular composition and spatial organization, with integration of these image-derived features with gene expression analyses.