An Active Contour Model Based on Adaptive Threshold for Extraction of Cerebral Vascular Structures.

Area of Science:

• Medical Imaging
• Computational Neuroscience
• Image Analysis

Background:

• Accurate cerebral vessel segmentation is crucial for clinical diagnosis and research.
• Automated segmentation is challenging due to complex vessel geometry and variable shapes.
• Time-of-flight Magnetic Resonance Angiography (TOF-MRA) data presents specific segmentation difficulties.

Purpose of the Study:

• To develop and evaluate a new active contour model (ACM) for segmenting cerebral vessels from TOF-MRA data.
• To improve the accuracy and completeness of brain vessel segmentation, particularly for thin vessels.
• To enhance the Dice Similarity Coefficient (DSC) compared to existing methods.

Main Methods:

• Implementation of a novel active contour model using the level-set method.
• The model's energy function integrates region intensity and boundary information with region, boundary, and penalty terms.
• Utilized a global threshold for thick vessels and a dynamic threshold for tiny vessels.
• Incorporated a boundary term for gradient-driven contour evolution and a penalty term to prevent reinitialization.

Main Results:

• The proposed ACM achieved a superior Dice Similarity Coefficient compared to global threshold and localized hybrid level-set methods.
• Successfully segmented entire cerebral vessel trees, including fine and intricate vessels.
• Demonstrated robust performance across 10 clinical brain datasets.

Conclusions:

• The novel active contour model offers a more accurate and comprehensive approach to cerebral vessel segmentation from TOF-MRA data.
• This method has the potential to significantly aid in clinical diagnosis and neurovascular research.
• The dynamic thresholding strategy effectively addresses the challenge of segmenting small-caliber vessels.