An artifact-robust, shape library-based algorithm for automatic segmentation of inner ear anatomy in...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Otolaryngology

Background:

• Cochlear implants (CIs) restore hearing but require precise programming.
• Image-guided CI programming (IGCIP) uses electrode position relative to cochlear anatomy.
• Pre-implantation CT scans are often needed for IGCIP due to artifacts in post-implantation scans.

Purpose of the Study:

• To develop an algorithm for automatic segmentation of intra-cochlear anatomy in post-implantation CT scans.
• To enable IGCIP using only post-implantation CT, overcoming the need for pre-implantation scans.

Main Methods:

• Developed an algorithm to automatically segment intra-cochlear anatomy in post-implantation CT scans.
• Identified the labyrinth using a library of surface maps and a shape model-based segmentation method.
• Tested the algorithm on 10 ears to evaluate accuracy.

Main Results:

• Achieved an overall mean error of 0.209 mm and a maximum error of 0.98 mm in segmenting intra-cochlear anatomy.
• Demonstrated accurate localization of intra-cochlear structures using post-implantation CT alone.

Conclusions:

• The developed algorithm is accurate enough for IGCIP strategies based solely on post-implantation CT scans.
• This technique expands the applicability of IGCIP to patients without pre-implantation CT data.