cDNA microarray image segmentation with an improved moving k-means clustering method

Different clustering based strategies have been proposed to increase the performance of image segmentation. However, due to complexity of chip preparing process, the real microarray image will contain artifacts, noises, and spots with different shapes, which result in these segmentation algorithms can't meet the satisfactory results. To overcome those drawbacks, this paper proposed an improved k-means clustering based algorithm to improve the segmentation accuracy rate. Firstly, an automatic contrast enhancement method is introduced to improve the image quality. Secondly, the maximum between-class variance gridding is conducted to separate the spots into sole areas. Then, we combine the k-means clustering algorithm with the moving k-means clustering method to gain a higher segmentation precision. In addition, an adjustable circle means is used for missing spots segmentation. Finally, intensive experiments are conducted on GEO and SMD data set. The results shows that the method presented in this paper is more accurate and robustness.