Segmentation of two dimensional electrophoresis gel image using the wavelet transform and the watershed transform

Abstract

Segmentation of two dimensional electrophoresis (2DE) gel image is a challenging task due to presence of nonlinear backgrounds, horizontal and vertical streaks, and irregular spots. The watershed method is a powerful tool for medical image segmentation, but it produces over-segmented results due to presence of noise and non-linearity. The solutions available in literature have failed to give satisfactory results in case of gel images. This paper presents a novel method for segmentation of 2DE gel images. The pitfalls of the watershed transform have been addressed through spot characterization in the wavelet domain. The wavelet transform is an important multi-scale analysis tool for the images. The proposed method utilizes the best features of both the watershed and the wavelet transforms in which connected maxima set corresponding to each watershed region has been introduced and computed in the wavelet domain. This allows us for accurate detection of the spots in each watershed region. Experimental results on the set of real gel images demonstrate that our method outperforms the commercialized software. Our method has also an advantage of single threshold parameter selection.