A shape-independent algorithm for fully-automated gridding of cDNA microarray images

In this paper, a fully-automated microarray gridding algorithm is presented. This algorithm contains the block finding in an image by using variable length Blackman window, image contrast enhancement based on the Otsu thresholding approach, and identification of image objects including spots and artifacts through the 8-connected labeling method. Furthermore, a shape-independent algorithm based on the area of pixels in each object is proposed for noise elimination. The final gridding is performed using a new method based on the constructed spot matrix and a refinement procedure is exploited to minimize all probable grid-line errors. The performance of the proposed algorithm is evaluated by five public datasets including the Swiss Institute of Bioinformatics (SIB), Joe DeRisi's individual tiff files (DeRisi), University of California, San Francisco (UCSF), Gene Expression Omnibus (GEO), and Stanford Microarray Database (SMD). The obtained results reveal that the proposed algorithm reaches a higher level of accuracy and stability against some restrictions in microarray images such as noise, artifacts, and irregularities regarding the shapes of spots in comparison with other state-of-the-art methods.