OBJECTIVE: A general method was developed to analyze and describe tree-like structures needed for evaluation of complex morphology, such as the cerebral vascular tree. Clinical application of the method in neurosurgery includes planning of the surgeon's intraoperative gestures. METHOD: We have developed a 3D skeletonization method adapted to tubular forms with symbolic description. This approach implements an iterative Dijkstra minimum cost spanning tree, allowing a branch-by-branch skeleton extraction. The proposed method was implemented using the laboratory software platform (ArtiMed). The 3D skeleton approach was tested on simulated data and preliminary trials on clinical datasets mainly based on magnetic resonance image acquisitions. RESULTS: A specific experimental evaluation plan was designed to test the skeletonization and symbolic description methods. Accuracy was tested by calculating the positioning error, and robustness was verified by comparing the results on a series of 18 rotations of the initial volume. Accuracy evaluation showed a Haussdorff's distance always smaller than 17 voxels and Dice's similarity coefficient greater than 70 %. CONCLUSION: Our method of symbolic description enables the analysis and interpretation of a vascular network obtained from angiographic images. The method provides a simplified representation of the network in the form of a skeleton, as well as a description of the corresponding information in a tree-like view.