Genes expression level quantification using a spot-based algorithmic pipeline

An efficient spot-based (SB) algorithmic pipeline of clustering, enhancement, and segmentation techniques was developed to quantify gene expression levels in microarray images. The SB-pipeline employed i/a griding procedure to locate spot-regions, ii/a clustering algorithm (enhanced fuzzy c- means or EnFCM) to roughly segment spots from background and estimate background noise and spot's center, iii/an adaptive histogram modification technique to accentuate spot's boundaries, and iv/a segmentation algorithm (Seeded Region Growing or SRG), to extract microarray spots' intensities. Extracted intensities were comparatively evaluated in term of Mean Absolute Error (MAE) against the MAGIC TOOL's SRG employing a dataset of 7 replicated microarray images (6400 spots each). MAE box-plots mean values were 0.254 and 0.630 for the SB-pipeline and the MAGIC TOOL respectively. Total processing times for the dataset evaluated (7 images) were 2100 seconds and 3410 seconds for the SB-pipeline and MAGIC TOOL respectively.