Whole Abdominal Wall Segmentation using Augmented Active Shape Models (AASM) with Multi-Atlas Label Fusion and Level...

Area of Science:

• Medical imaging analysis
• Computational anatomy
• Radiology

Background:

• The abdominal wall is crucial for anatomical structure and compartmentalization.
• Accurate segmentation of the abdominal wall in computed tomography (CT) scans is challenging due to anatomical complexity.
• Existing segmentation methods struggle with the variations and intricacies of the abdominal wall and surrounding tissues.

Purpose of the Study:

• To develop a robust and automated method for segmenting the abdominal wall's inner and outer surfaces on CT scans.
• To improve the accuracy of subcutaneous and visceral fat volume measurements.
• To present a novel algorithm combining active shape models, multi-atlas label fusion, and level sets.

Main Methods:

• A slice-wise augmented active shape model (AASM) approach was developed.
• Multi-atlas label fusion (MALF) and level set (LS) techniques were integrated into the active shape model (ASM) framework.
• The AASM approach optimizes landmark updates within complex anatomical contexts, validated on 20 CT scans (184 slices).

Main Results:

• The proposed AASM approach significantly reduced the Hausdorff distance compared to individual ASM, MALF, and LS methods.
• Automated segmentation enabled accurate subcutaneous and visceral fat measurements, showing high correlation with manual segmentation.
• The combined algorithm demonstrated superior performance in segmenting the abdominal wall.

Conclusions:

• The developed AASM algorithm offers a robust and accurate solution for abdominal wall segmentation in CT imaging.
• This method facilitates precise quantification of visceral and subcutaneous fat volumes, aiding clinical assessments.
• The study presents the first generic algorithm integrating ASM, MALF, and LS for automated fat volume capture.