A clinically motivated 2-fold framework for quantifying and classifying immunohistochemically stained specimens

  • Authors:

  • Bonnie Hall;Wenjin Chen;Michael Reiss;David J. Foran

  • Affiliations:

  • Center for Biomedical Imaging and Informatics and Graduate School of the Biomedical Sciences and The Cancer Institute of New Jersey, UMDNJ, Robert Wood Johnson Medical School, New Brunswick, NJ;Center for Biomedical Imaging and Informatics and The Cancer Institute of New Jersey, UMDNJ, Robert Wood Johnson Medical School, New Brunswick, NJ;The Cancer Institute of New Jersey, UMDNJ, Robert Wood Johnson Medical School and Dept. of Internal Medicine and Dept. of Molecular Genetics, Microbiology and Immunology, New Brunswick, NJ;Center for Biomedical Imaging and Informatics and Graduate School of the Biomedical Sciences and The Cancer Institute of New Jersey, UMDNJ, Robert Wood Johnson Medical School, New Brunswick, NJ

  • Venue:
  • MICCAI'07 Proceedings of the 10th international conference on Medical image computing and computer-assisted intervention
  • Year:
  • 2007

Quantified Score

Hi-index 0.03

Visualization

Abstract

Motivated by the current limitations of automated quantitative image analysis in discriminating among intracellular immunohistochemical (IHC) staining patterns, this paper presents a two-fold approach for IHC characterization that utilizes both the protein stain information and the surrounding tissue architecture. Through the use of a color unmixing algorithm, stained tissue sections are automatically decomposed into the IHC stain, which visualizes the target protein, and the counterstain which provides an objective indication of the underlying histologic architecture. Feature measures are subsequently extracted from both staining planes. In order to characterize the IHC expression pattern, this approach exploits the use of a non-traditional feature based on textons. Novel biologically motivated filter banks are introduced in order to derive texture signatures for different IHC staining patterns. Systematic experiments using this approach were used to classify breast cancer tissue microarrays which had been previously prepared using immuno-targeted nuclear, cytoplasmic, and membrane stains.