This work introduces a novel level-set based scheme for the segmentation of protein spots in 2D-GE images. The proposed scheme incorporates a protein spot detection stage based on the identification of local intensity minima in order to allow the separation of overlapping spots. Moreover, histogram adaptation and morphological operators are applied for the identification of faint spots. A segmentation stage follows, where a level-set function is initialized based on local intensity maxima and evolves according to energy terms derived in the detection stage. The experimental evaluation demonstrates that the proposed scheme generates more plausible spot boundaries than PDQuest 8.0.1 and Melanie 7 and outperforms both software packages in terms of segmentation performance.