The MIDAS Journal logo

The MIDAS Journal


Biomarker Detection in Whole Slide Imaging based on Statistical Color Models

Shu, Jie, Qiu, Guoping, Ilyas, Mohammad, Kaye, Philip
University of Nottingham
The MIDAS Journal logo

Please use this identifier to cite or link to this publication:
New: Prefer using the following doi:
Published in The MIDAS Journal - MICCAI 2010 Workshop: Computational Imaging Biomarkers for Tumors (CIBT).
Submitted by Guoping Qiu on 2010-09-03 04:42:25.

This paper presents a technique for immunostaining biomarker detection in digital slides. We treat immunostaining detection as a color image analysis problem and build statistical color models using a large number of labeled positive and negative immunostaining pixels. We have implemented the statistical models in different color spaces and show that the opponent chromaticity signals effectively characterize the color distributions of the immunostaining biomarkers and that the luminance is an unreliable and distractive signal. We have applied the technique to the detection of positive P53 immunostaining in digital slides of oesophagitis and colorectal biopsies. We present experimental results and show that the technique can achieve a biomarker detection rate of over 98% with 5% false positives.