Gene expression-guided selection of histopathology image features

Histopathology imaging and gene expression profiling are two fundamental investigative techniques which allow the analysis of biological specimens from different perspectives. Given their apparent divergence in data representation, they are usually used separately, being connected only at the higher levels of data analysis. In this work we demonstrate how gene expression can be used directly for guiding the selection of prognostically-relevant imaging features. Our method is applied to the analysis of a breast cancer data set, but is not limited to this pathology.