Dots Detection in HER2 FISH Images Based on Alternative Color Spaces

Abstract

This article presents an algorithm for detection of HER2 gene and chromosome centromere - CEN17 signals in FISH images. FISH is a popular fluorescence staining method which allows the evaluation the degree of HER2 gene amplification on breast cancer tissues. The FISH test allows the choice of an appropriate treatment. To ensure the best discrimination of the spots from the non-spots elements (i.e. a stroma) the algorithm was based on the color spaces transformation. The main point of this algorithm is an automatic red and green spots detection using transformations into different color spaces. Quality-identification tests were performed by using one-class SVM. The selection of an optimal color space for identification of the spots was completed using automatic Fisher feature selection. The results of the experiments are promising and show good effect of color transformation application on genes recognition system. Detailed results are presented.