A comparison of medical image segmentation methods for cerebral aneurysm computational hemodynamics

Patient-specific hemodynamic technology has been applied in clinical applications. However, the process of vessel segmentation was insufficiently validated. In order to confirm the accuracy of medical image segmentation methods, 13 image segmentation methods are introduced in this study to compare the results of cerebral-vascular aneurysms and its parent arteries from three-dimensional computed tomography (3D CT) images. This study indicates that the volume of the aneurysm models can reach difference of 11% with different segmentation methods under the same intensity threshold. The same segmentation methods under different intensity ranges can cause a volume change of up to 18%. The segmentation method also influences the local geometric shapes of the aneurysms. Some segmentation methods change subtle aspects of the anatomical shapes, which significantly influences the hemodynamic analysis and clinical decision. Computational hemodynamic simulation is performed by using the geometric results from segmentation. The hemodynamic characters; i.e. energy loss, are found to have a maximum of 34.8% in difference from segmentation. The results indicate that validation will be an essential process in the confirmation of the segmentation quality of patient-specific cerebral-vascular hemodynamic analysis.