Atlas-based segmentation of cochlear microstructures in cone beam CT.

Area of Science:

• Medical Imaging
• Anatomical Segmentation
• Surgical Simulation

Background:

• Accurate segmentation of cochlear microstructures is crucial for surgical planning.
• Clinical cone beam computed tomography (CBCT) offers detailed temporal bone imaging.
• Existing segmentation methods may lack automation and precision for complex anatomical structures.

Purpose of the Study:

• To develop an automated segmentation method for key cochlear microstructures (scala tympani, scala vestibuli, modiolus, mid-modiolus, round window membrane).
• To enable the use of segmented data in surgical simulation software.
• To support preoperative surgical evaluation and planning.

Main Methods:

• Utilized the OpenEar (OE) Library with registered CBCT and 3D micro-slicing images.
• Created a cochlear microstructure atlas from OE data.
• Applied affine registration of the atlas to automatically segment structures in other OE CBCT images.
• Validated segmentations using quantitative metrics (DICE, Hausdorff distance) and visual review.

Main Results:

• Achieved average DICE scores of 0.77 for scala tympani and 0.74 for scala vestibuli.
• Reported average Hausdorff distances below 0.12 mm for both scalae.
• Demonstrated accurate segmentation of the round window membrane and mid-modiolar axes with minimal mean distance and angular deviation.
• Visual assessment confirmed segmentation accuracy comparable to expert manual tracing.

Conclusions:

• An atlas-based, automated segmentation approach using 3D micro-slicing and affine registration is effective for cochlear microstructures.
• This method successfully segments temporal bone anatomy for surgical simulation.
• The approach shows potential for enhancing pre-surgical planning and rehearsal.