TACOMA

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Donghui Yan*+ and Pei Wang and Tim Randolph
• Companies: Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
• Keywords: Tissue microarray images ; computational analysis ; image scoring ; Random Forests

Abstract

Tissue microarray (TMA) technology allows one to evaluate large numbers of immunohistochemically-stained tissue images for the presence of a protein marker and has been used increasingly often for the evaluation of diagnostic tests and validation of biomarkers. In response to concerns about the subjectivity and variability of pathologist-based TMA evaluation, we present the TACOMA algorithm to automatically score TMAs in an efficient and objective manner. The texture regularity hidden in the TMAs are effectively captured by a summary statistics. Guided by a few `representative' image patches, TACOMA is able to focus on biologically relevant features. Experiments with TMA images (for ER biomarker in breast cancer samples) show that TACOMA is comparable to pathologists in terms of accuracy and reproducibility, moreover, it reveals salient features in an image most relevant to scoring.