Analyzing the structure of the left atrium can provide precious insights into the pathology of atrial fibrillation, eventually resulting in optimization of treatment plans. In this paper, an interactive and patient-specific method is presented to segment the left atrial endocardium, the left atrial epicardium and measure the left atrial wall thickness from cardiac computed tomography images. A region growing algorithm was adapted to segment the left atrial endocardium, whereas the left atrial epicardium was segmented indirectly: a marker-controlled geodesic active contour model was defined on its surrounding environment. The results of the left atrial wall thickness were then mapped onto meshes generated from the endocardium segmentation. We tested our pipeline on 10 datasets as a part of the STACOM 2016 Left Atrial Wall Segmentation Challenge and we compared our method with manual segmentation. Aimed at facilitating the segmentation of the left atrial thin-wall structure, this pipeline is partially implemented in MUSIC software for clinical use. The expertise of clinicians can be added through the choice of specific parameters for each patient, although this remains optional owing to the robustness of the approach.