Segmentation of giant cerebral aneurysms using a multilevel object detection scheme based on lattice Boltzmann method

Segmentation of giant cerebral aneurysms composed of lumen and thrombus in computed tomography angiography remains a challenge for image processing community. We propose a multilevel object detection scheme based on lattice Boltzmann method (LBM) to tackle this problem. The method consists of first extracting the lumen using the LBM with some parameter configuration, and then detecting thrombus with the aid of a different LBM configuration, and finally expanding a disc to refine the shape of the thrombus. The results show that with the proposed method both lumen and thrombus can be well segmented, demonstrating the promising potentiality of this LBM-based multilevel detection approach for clinical applications.