The diagnostic of atherosclerosis is a major stake in order to understand the evolution and the consequences of this disease. Towards the fusion of two promising imaging techniques, IVUS and OCT, the goal of this thesis is to detect automatically the corresponding images in both modalities. Our method is based on the characterization of the lumen shape by a set of features. We compute this features, such as the PCA Eigenvalues the perimeter and the area of the lumen after a segmentation in both IVUS and OCT data. The segmentation of the lumen is performed by a snake algorithm using a gradient vector flow as external force. We test our approaches on pig and human pullbacks and the results are somehow disappointing. The method is heavily dependant on the quality of the segmentation, such that the results can be improved by improving the segmentation. The computation of additionnal features should also improve the efficiency of our approach.