Multi-atlas active contour segmentation method using template optimization algorithm.

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Accurate brain image segmentation is crucial for diagnosis, treatment planning, and 3D reconstruction.
• Manual segmentation is time-consuming and subjective, necessitating automated or semi-automated methods.

Purpose of the Study:

• To develop a semi-automatic brain image segmentation method combining multi-atlas registration and an active contour model (ACM).
• To improve the accuracy and efficiency of brain image segmentation for clinical applications.

Main Methods:

• A multi-atlas registration method with affine transformation was employed to obtain prior shape information.
• A template optimization algorithm was used to reduce registration errors and generate an initial active contour (IAC).
• An active contour model (ACM) was utilized for the final segmentation of target brain tissues.

Main Results:

• The method achieved high accuracy on publicly available MR datasets (IBSR and MRBrainS13).
• Average Dice similarity coefficients of 0.927±0.014 (thalamus) and 0.853±0.03 (gray matter) were obtained.
• Efficient registration times were achieved using affine transformation, with average times of 255s (IBSR) and 409s (MRBrainS13).

Conclusions:

• The proposed semi-automatic method effectively segments brain tissues with improved accuracy and reduced registration time.
• The combination of multi-atlas registration and ACM offers a robust solution for brain image segmentation.
• The method provides smooth, continuous contours essential for clinical analysis and 3D reconstruction.