Detection of Spots in 2-D Electrophoresis Gels by Symmetry Features [chapter]

Martin Persson, Josef Bigun
2005 Lecture Notes in Computer Science  
We have implemented an algorithm for detection and segmentation of protein spots in 2-D gel electrophoresis images using symmetry derivative features computed using low level image processing operations. The implementation was compared with a previously published Watershed segmentation and a commercial software. Our algorithm was found to yield segmentation results that were either better than or comparable to the other solutions while having fewer free parameters and a low computational cost.
more » ... ummary Two-dimensional gel electrophoresis (2-DE) is a major workhorse in proteomics. 2-DE data comes as spot maps containing a vast number of proteins, requiring automatic image processing for efficient analysis. Quantification of individual proteins and tracing changes in expression between gels require accurate spot detection. Existing spot detection algorithms often require user intervention for setting free parameters and time consuming morphological post processing. We approach the problem by using a set of computationally cheap and robust symmetry derivative features and minimal post processing. A feed forward neural network is used to find decision boundaries in the feature space. The neural network is trained with features extracted from manually segmented 2-DE images. Classification performance is compared with the published non-commercial algorithm of Bettens [1] and one commercial 2DE image analysis program, ImageMaster™ 2D Platinum v5.0 (GE Healthcare, formerly Amersham Biosciences). The result, presented as ROC curves, show that we perform at least as well as both Bettens and Imagemaster in terms of spot detection and segmentation, while using fewer free parameters, and a limited amount of computational resources. Originality and Contribution We propose a set of symmetry derivative features [2, 3] to be used in automatic segmentation of 2DE gel images with minimal post-processing. Symmetry derivatives give immediate information on local shape that otherwise requires time consuming regional processing. In addition to achieving better segmentation performance, this moves the focus of the problem from post processing to basic signal processing.
doi:10.1007/11551188_47 fatcat:ftnxylthu5bxvmet27ot4m6sbe