Detecting & Analyzing Of Gene Chip Image Signal Based On Expectation Maximization Algorithm

Gene chip, a new applied biologic technology, has been discovered in 1980s. With the phenomenal development of gene chip technology, it has become a standard tool in many genomics research laboratories.Gene chip is used for the concurrent identification of thousands of genes in the field of bioinformatics. By finding the location of the spots, extracting and calculating the spots'average or median of fluorescent intensity, specific gene data can be obtained and gene expression can be achieved. Thus, image processing and data analysis are two important steps of gene chip experiments.Based on the digital image processing and the characteristic of gene chip image, this thesis focuses on four steps, gene chip image preprocessing, spots gridding, image segmentation and data extraction and expression. By summarizing and carrying out the present methods, a new technique is introduced in this thesis. Details of this thesis are as follows.First, gene chip image preprocessing is introduced, in which image filtering plays an important step in the gene chip experiments. In this section, we discuss several general filtering methods. Then a new method based on the Offset Vector Field is introduced. Simulations and comparison show that this new method can detect the edge of spots eliminate the noises and remain the targets of image.Second, spots gridding is to detect the location of spots in the image. Based on the characteristic of gene chip image, the projective method is adopted for gridding. Experiments indicate that it is stable and accurate.Third, image segmentation is a critical step in gene chip image processing. In this section, general methods of image segmentation are discussed and simulated. Then a new method based on the Expectation Maximization Algorithm is presented in detail. After simulations, it is showed that this method is accurate, stable and has no restriction on shape, size and intensity of spots.Finally, after extracting the spots information, the special gene data, such as spot intensity, background intensity and so on, can be obtained, and then data expression and analysis is carried out using the MA plot.