摘要
Mediastinum segmentation, as well as lung segmentation, has numerous applications in medical imaging and computer-aided diagnosis. This paper presents an anatomy-based mediastinum segmentation algorithm in CT images. The proposed method actualizes 3D mediastinum segmentation by taking the context-contour of proximal 2D segment results into account. Margin points inspired by anatomical knowledge are proposed to roughly describe 2D mediastinal area and are extracted based on its specific anatomic features. Fast marching method is then designed to extract the path to avoid the edge of mediastinal area crossing opacified vessel and spines. The experiment has performed on 19 CT scan data for pulmonary embolism detection, and the results presented that 19/19 segmented 3D mediastinum results satisfied experiment radiologists, and the proposed method performances well in both accuracy and efficiency.
- 出版日期2011