Deformable abdominal wall segmentation

• Main Author: Song, Menglu.
• Other Authors: Lin Zhiping
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Bioengineering
• Online Access: http://hdl.handle.net/10356/54361
• Institution: Nanyang Technological University

This report presents a semi-automatic method for abdominal wall segmentation from 3D Computed Tomography (CT) images using Active Shape Model (ASM) as well as Extreme Learning Machine (ELM). As a top-down approach, ASM builds a shape model from a set of sample images where the abdominal wall has been annotated manually. By iteratively adjusting the pose and shape parameters of the model, a best fit to a new image can be found such that the model can describe abdominal wall as accurately as possible. On the other hand, ELM as a bottom-up approach is adopted to distinguish wall voxels with other non-wall voxels. Partial wall can be extracted to deform the model in the fitting process of ASM. The experimental result shows that the proposed ASM and ELM method is capable of delivering reasonably good performance for abdominal wall segmentation. To the author’s best knowledge, this is the first work that applies ASM and ELM for abdominal wall segmentation.