Hybrid Spot Segmentation in Four-Channel Microarray Genotyping Image Data

In this paper we present a novel hybrid algorithm for spot segmentation in four-channel genotyping microarray images. A new four-dimensional clustering approach for fully-automated spot segmentation is proposed, along with a new iterative method to automatically identify the number of clusters in a single-spot area. A spatial shape detection step is simultaneously applied, which assists a nonlinear diffusion filtering step in detecting the connected objects, while a spot masking step prevents various noise types from misleading the spot extraction algorithm. The developed analysis system successfully handles various four-channel as well as traditional two-channel microarray image data obtained from different sources. The platform is tested on 34 microarray data sets. The segmentation algorithm accurately segments donut-shaped spots, as well as irregular spot shapes, spots with intensity variations and different noise types effectively. The data extracted from the samples is fed to genotyping algorithms with the results achieving high accuracy rates of up to 99.8% with a call rate of 90.8%