An Unsupervised and Fully-Automated Image Analysis Method for cDNA Microarrays

Microarray gene expression image analysis is a labor-intensive task and requires human intervention since microarray images are contaminated with noise and artifacts while spots are often poorly contrasted and ill-defined. The analysis is divided into two main stages: Gridding and Spot-Segmentation. In this paper, an original, unsupervised and fullyautomated approach to gridding and spot-segmenting microarray images, which is based on two genetic algorithms, is presented. The first genetic algorithm determines the optimal grid while the second one determines, in parallel, the boundaries of multiple spots. Experiments on 16-bit microarray images show that the proposed method is effective and achieves more accurate gridding and spot-segmentation results in comparison with existing methods.

• Type: Conference
• Year: 2007
• Where: CBMS
• Authors: Eleni Zacharia, Dimitrios E. Maroulis
• Keywords: CBMS 2007, Genetic Algorithm, Medical Imaging, Microarray Gene Expression, Microarray Images