Automated cDNA Microarray Segmentation using Independent Component Analysis Algorithm
There is a most useful method in order to simultaneously study thousands of gene expression levels of a simple experiment, called DNA microarray. The average value of fluorescent intensity can be calculated through a microarray experiment. The calculated intensity values are close to the expression levels of a specific gene. Therefore, the appropriate determination of every spot position (specific gene) will lead to the accurate measurement of the intensity amount, and as a result, accurate classification of normal and abnormal gene expression levels within the microarray image. In this paper, first, a preprocessing step is performed in order to cancel the noise and artifacts in DNA microarray images using the nonlinear diffusion filtering method. Then, the coordinate center of each spot is determined utilizing mathematical morphology operations. Finally, pixel classification is performed using the independent component analysis (ICA) algorithm. Performance of the proposed algorithm has been assessed on the microarray images of the Stanford Microarray Database (SMD). Realization results illustrate that the classification accuracy of the proposed algorithm of the noisy microarray cells is close to 98%, while this amount is equal to 100% for noiseless cells.
Gene expression, Microarray, Independent components, Noise