Automated Segmentation of Microarray Spots Using Fuzzy Clustering Approaches

Microarray imaging is now widely used to monitor the activities of thousands of genes simultaneously in biological samples. While there are a number of methods in use for the quantification of microarray images, barriers still exist towards its feasibility for clinical use. Among them, automated spot segmentation is critical for accurate and high throughput measurements of gene expression levels from a hybridization experiment. We introduce clustering based segmentation approaches such as fuzzy c-means clustering for this purpose. The red and green intensity values from the cy3 and Cy5 hybridization images are used as features to cluster each pixel into foreground and background. The proposed approaches overcome of the difficulty of most existing segmentation methods that do not consider the variable shape of the spots and the use of spectral correlations. The proposed algorithms have been tested on a variety of microarray spots, demonstrating their superior performance