An Improved Microarray Image Analysis Architecture Using Mathematical Morphology

DNA microarrays are now widely used to measure gene expression levels of healthy and cancerous cells. To allow further experiment for drug development to treat cancer, colour intensity from images of microarray spots need to be extracted as accurate as possible. The intensity extraction requires pre-requisite analysis stages including noise removal, and followed by location gridding. However, it remains as a challenging task for microarray analysis due to the variation of noise that infested the images. In this study, microarray analysis architecture using mathematical morphology was proposed, namely Mathematical Morphology Microarray Image Analysis (MaMIA).Firstly, in denoising stage, noise identification is conducted to identify and reverse the noise. Next, combinations of mathematical morphology were applied to the microarray and its pixel derivatives during the gridding stage. Raw microarrays used by MaMIA are available at Stanford Microarray Database (SMD), Gene Expression Omnibus (GEO) and from a dilution experiment (DILN). From comparisons with previous existing architectures, Optimal Multilevel Thresholding (OMTG) and Automated Robust MicroArray Data Analysis (ARMADA), MaMIA have proven to efficiently remove noise with highest value, 81.6657dB for Peak Signal to Noise Ratio (PSNR) and success identification of spots in cases of noises; with highest gridding accuracy level of 98.34%.Overall processing time, MaMIA architecture can perform gridding in less than 22 seconds, fastest as compared to its contender. This research have revealed the potential of analysing microarray by mainly using mathematical morphology operation, either applied on microarray or its pixel derivative.