MIA: An Effective and Robust Microarray Image Analysis System with Unstructured Information Management Architecture

The proposed microarray image analysis (MIA) system is designed to analyze microarray slide images in a fully automatic manner. This system identifies and rectifies tilted slides, discovers block boundaries, generates gridlines, recognizes spots, and finally extracts the accurate spot intensity values from the two image channels (red and green) in a microarray slide. The red-to-green intensity ratio of a spot represents the gene expression level in the specimen. Our experimental results demonstrate the effectiveness and robustness of the proposed system. Further, the MIA system is tightly integrated with the component-based Unstructured Information Management Architecture (UIMA) which is an open source platform for the analysis of unstructured data (e.g. images) and is developed by IBM. With UIMA, we can easily apply various analysis algorithms on data by simply plugging analysis components into the system. Further, the analysis results at each analyzing step are attached to the data object as its annotations. The major contribution of this paper is that we design a microarray image analysis system which provides users a convenient manner to automatically analyze slide images and acquire accurate gene expression data from microarray slides. Also, the proposed MIA system, which is based on UIMA, provides a flexible, scalable, and extensible environment for users to perform various analysis tasks on microarray slide images.