Quantitative analysis of images in erythropoietin doping control

Abstract

The software system GASepo has been developed as a tool to visualise and analyse doping (searching for performance-enhancing drugs in sports) with recombinant erythropoietin (rEpo). Digital images derived from the separation of Epo isoforms in gels by iso-electric focusing followed by double blotting and chemiluminescence detection contain spots (bands) of characteristic shape and positions. For Epo doping control, these have to be analysed and evaluated. A relevant element of the analysis is calculation of the reference cutoff-line (COL) to which the doping positivity criterion is related. Based on analysis of the previous method used, a novel method for the COL calculation was developed and validated. The methodology was based on generation of a partition of the image (lane) being processed into a system of adjacent subimages (blocks) and quantitative characterisation of intensity variability within these blocks by a suitably defined measure. The coordinate of the proper COL position, separating the image part with recombinant Epo bands from the homogeneous background, was calculated by minimisation of the difference function of the measures for two adjacent image blocks. The proposed method was tested on real Epo images and validated on a synthetic phantom of the Epo image. The deviations of the COL position are limited by a 7 pixel-wide corridor. For the cases of noise with standard deviations 2300 and 2600, the deviations did not exceed 2 pixels over the whole range of the lane width values. Testing the method on 50 real Epo images originating from different doping-control laboratories worldwide showed its robust behaviour in Epo images.