A hybrid geometric-statistical deformable model for automated 3-D segmentation in brain MRI

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• Accurate brain tissue segmentation is crucial for diagnosing neurological disorders.
• Current segmentation methods face challenges with complex anatomical structures and multi-sequence MRI data.

Purpose of the Study:

• To develop a novel 3-D deformable model for accurate and robust automated brain MRI segmentation.
• To integrate image edge geometry and voxel statistical homogeneity for improved segmentation accuracy.

Main Methods:

• An edge-based geodesic active contour model was employed for segmentation.
• A novel hybrid geometric-statistical feature was developed to regularize contour convergence.
• The method was validated on simulated and clinical brain MRI scans across multiple sequences (T1, T2, PD).

Main Results:

• The proposed method demonstrated accurate and robust segmentation of brain tissues.
• Significant improvements in Dice similarity indexes were achieved for white matter (8.55%) and gray matter (10.18%) compared to a state-of-the-art method.
• The approach effectively extracts complex anatomical structures.

Conclusions:

• The novel 3-D deformable model offers a significant advancement in automated brain MRI segmentation.
• The hybrid geometric-statistical feature enhances segmentation accuracy and robustness, particularly for multi-sequence data.
• This method holds promise for clinical applications in neuroimaging analysis.