Speech Title: Enhanced approach for detection and segmentation of protein spots on 2D gel electrophoresis images
Aims: Two-dimensional gel electrophoresis (2D-GE) is a powerful and well-established
tool for separating complex protein mixtures based on their isoelectric points and their
molecular weights. 2D-GE images, the digital output of the 2D-GE technique, may
contain up to ~10.000 protein spots. The analysis of 2D-GE images is a challenging task
consisting of spot detection, segmentation and quantification. Available software
packages and techniques have the disadvantage of detecting a high number of spurious
(false positives) spots, while missing multiple actual spots (false negatives). The
proposed approach is effective in spot detection, segmentation and quantification and
outperforms the compared state-of-the- art software packages and techniques.
Methods: The proposed methodology consists of a multi-thresholding scheme applied on overlapping regions of 2D-GE images, a custom grow-cut segmentation algorithm, as well as a region growing scheme and morphological operators.
Results: The proposed approach has been evaluated on real and synthetic 2D-GE
images. Results demonstrate that it successfully addresses the task of spot detection,
segmentation and quantification. Additionally, comparative evaluation with
state-of-the-art software packages and techniques shows that the proposed approach
outperforms the other methods in spot detection as it achieves: i) the highest precision
value (lowest number of false positive spots), ii) a high sensitivity value (low number of
false negative spots), iii) the highest F-measure value, i.e. the harmonic mean of
precision and sensitivity, and iv) more plausible spot boundaries in both real and
synthetic images. Furthermore, a quantitative comparative evaluation for the task of
segmentation of synthetic 2D-GE images showed that the proposed approach achieved a
high Volumetric Overlap (VO) and the lowest Volumetric Error (VE) and Volumetric
Overlap Error (VOE) values, demonstrating its effectiveness compared to the examined
software packages and techniques.
Conclusions: In this work, an original approach for segmenting 2D-GE images is
presented. The experimental evaluation shows that the proposed approach provides
more accurate results than the available state-of-the-art software packages and
techniques, resulting in improved quantification of protein expression levels and as a
consequence enhancing the reliability of the extracted biological conclusions.