Fast approximation for joint optimization of segmentation, shape, and location priors, and its application in...

Area of Science:

• Medical imaging analysis
• Computational anatomy
• Image segmentation

Background:

• Inter-subject variability in organ location poses challenges for accurate segmentation.
• Existing exact optimization methods for segmentation and shape priors are computationally intensive.

Purpose of the Study:

• To develop a fast approximation for joint optimization of segmentation, shape, and location priors.
• To overcome computational complexity in organ segmentation, specifically for gallbladders in non-contrast CT volumes.

Main Methods:

• Simultaneous optimization of segmentation, shape, and location priors using a branch-and-bound method after spatial standardization.
• Fast approximation achieved by sampling in eigenshape space and efficient lower bound evaluation.

Main Results:

• Joint optimization including translation significantly improved segmentation performance.
• The proposed method achieved a Jaccard index of 0.623 for gallbladder segmentation, comparable to state-of-the-art.
• The algorithm demonstrated good computational efficiency, suitable for personal computer execution.

Conclusions:

• Joint optimization of segmentation, shape, and location priors is effective for gallbladder segmentation.
• The proposed approximation method offers high computational efficiency.
• This approach addresses inter-subject variability in organ location for improved medical image analysis.