Mesenteric vasculature-guided small bowel segmentation on 3-D CT.

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Automatic small bowel segmentation is crucial for diagnosing small bowel pathology via computer-aided diagnosis, aiding in visualization, tumor detection, and preoperative planning.
• Direct segmentation on computed tomography (CT) scans is difficult due to low contrast, high small bowel tortuosity, and proximity to other abdominal organs.

Purpose of the Study:

• To propose a novel mesenteric vasculature map-guided method for segmenting the small bowel on high-resolution CT angiography scans.
• To leverage the anatomical relationship between mesenteric vasculature and the small bowel for improved segmentation accuracy.

Main Methods:

• Segmentation of major mesenteric arteries using a vessel tracing method with multi-linear subspace vessel models and Bayesian inference.
• Identification of mesenteric vessels via multi-view, multi-scale vesselness enhancement filters, classifying vessels connected to the superior mesenteric artery.
• Construction of a mesenteric vasculature map by linking bifurcation points, followed by small bowel segmentation using the map and fuzzy connectedness.

Main Results:

• The proposed method achieved 82.5% volume overlap accuracy compared to a manually labeled reference standard.
• Evaluation was performed on 11 abdominal CT scans from patients suspected of carcinoid tumors.

Conclusions:

• The mesenteric vasculature map-guided approach demonstrates feasibility for segmenting the small bowel on CT scans.
• This method shows promise as a reliable tool for computer-aided diagnosis of small bowel pathologies.