We propose a novel method, the adaptive local window, for improving level set\nsegmentation technique. The window is estimated separately for each contour\npoint, over iterations of the segmentation process, and for each individual\nobject. Our method considers the object scale, the spatial texture, and changes\nof the energy functional over iterations. Global and local statistics are\nconsidered by calculating several gray level co-occurrence matrices. We\ndemonstrate the capabilities of the method in the domain of medical imaging for\nsegmenting 233 images with liver lesions. To illustrate the strength of our\nmethod, those images were obtained by either Computed Tomography or Magnetic\nResonance Imaging. Moreover, we analyzed images using three different energy\nmodels. We compare our method to a global level set segmentation and to local\nframework that uses predefined fixed-size square windows. The results indicate\nthat our proposed method outperforms the other methods in terms of agreement\nwith the manual marking and dependence on contour initialization or the energy\nmodel used. In case of complex lesions, such as low contrast lesions,\nheterogeneous lesions, or lesions with a noisy background, our method shows\nsignificantly better segmentation with an improvement of 0.25+- 0.13 in Dice\nsimilarity coefficient, compared with state of the art fixed-size local windows\n(Wilcoxon, p < 0.001).\n