A Fully Automatic Approach for Enhancement of Microarray Images

Abstract

DNA microarray technology has promised a very accelerating research inclination in recent years. There are numerous applications of this technology, including clinical diagnosis and treatment, drug design and discovery, tumor detection, and in the environmental health research. Enhancement is the major pre-processing step in microarray image analysis. Microarray images when corrupted with noise may drastically affect the subsequent stages of image analysis and finally affects gene expression profile. This paper presents an approach that achieves an automated way for applying mathematical morphology for the enhancement of microarray images. White and black top-hat transform is performed to denoise the image. A threshold is estimated which is dependent on image characteristics to remove artifacts present in the image. Experiments on Stanford, TBDB and UNC database illustrate robustness of the proposed approach in the presence of noise, artifacts and weakly expressed spots. Experimental results and analysis illustrates the performance of the proposed method with the contemporary methods discussed in the literature. Index Terms-microarray, morphology and white & black top-hat transform.