Automated atlas-based segmentation for skull base surgical planning.

Area of Science:

• Neurosurgery
• Medical Imaging
• Computational Anatomy

Background:

• Computational surgical planning tools are crucial for developing novel skull base surgical approaches.
• Automated skull base segmentation is needed to enhance the usability of surgical planning software.
• This work addresses the need for improved accuracy and efficiency in skull base surgical planning.

Purpose of the Study:

• To design and validate an atlas-based automated segmentation algorithm for skull base structures.
• To improve the usability of skull base surgical planning software through automated segmentation.
• To enhance safety and patient outcomes in skull base surgery.

Main Methods:

• A synthetic CT template was created and skull base structures were manually segmented to form a reference atlas.
• Advanced Normalization Tools and Elastix deformable registration were used to register the atlas to 30 individual image sets.
• Dice coefficient, average Hausdorff distance, and clinical usability scoring were employed for validation.

Main Results:

• The mean average Hausdorff distance for all segmented structures was below 2 mm.
• The average Dice coefficient for larger structures (>2.5 mL) was 0.73, with smaller structures showing coefficients below 0.7.
• Clinical usability scoring indicated acceptable effort for most structures, with minor exceptions requiring more significant corrections.

Conclusions:

• Open-source algorithms like Elastix enable accurate automated atlas-based segmentation of skull base structures.
• The proposed atlas and CT template offer acceptable clinical accuracy with minimal corrections.
• The release of a public CT template and segmentation atlas facilitates general use in skull base surgical planning.