| Microarray imaging is a powerful technique developed in functional genomics for large-scale gene expression analysis. This technology is now widely used for drug discovery, disease diagnosis and prognosis. Despite the widespread application of microarray imaging in life sciences, barriers still exist regarding its reliability and reproducibility for clinical use. A major critical problem lies in the accurate spot segmentation and quantification of gene expression level from the microarray spot images.; We introduce fuzzy clustering based segmentation approaches for this purpose. This approach overcomes the difficulties of most existing segmentation methods that neither consider the variable shape of the spots nor use the two-channel spectral information. In addition, three statistical criteria are compared in measuring gene expression levels, which indicates that a new unbiased statistic is more suitable for quantifying the gene expression level. The proposed algorithms demonstrate better performance on both simulated and real microarray images. |
