Automatic gridding of microarray images based on spatial constrained K-means and Voronoi diagrams

Images from complementary DNA (cDNA) microarrays need to be processed automatically due to the huge amount of information that they provide. In addition, automatic processing is also required to implement batch processes able to manage large image databases. Most of existing softwares for microarray image processing are semiautomatic, and they usually need user intervention to select several parameters such as positional marks on the grids, or to correct the results of different stages of the automatic processing. On the other hand, many of the available automatic algorithms fail when dealing with rotated images or misaligned grids. In this work, a novel automatic algorithm for cDNA image gridding based on spatial constrained K-means and Voronoi diagrams is presented. The proposed algorithm consists of several steps, viz., image denoising by means of median filtering, spot segmentation using Canny edge detector and morphological reconstruction, and gridding based on spatial constrained K-means and Voronoi diagrams computation. The performance of the algorithm was evaluated on microarray images from public databases yielding promising results. The algorithm was compared with other existing methods and it shows to be more robust to rotations and misalignments of the grids. Red de Universidades con Carreras en Informática (RedUNCI)