Automatic segmentation of intracochlear anatomy in conventional CT

Area of Science:

• Medical Imaging
• Otolaryngology
• Biomedical Engineering

Background:

• Profound hearing loss treatment often involves cochlear implant surgery.
• Optimal electrode placement in the scala tympani is linked to better hearing restoration.
• Conventional imaging struggles to visualize thin intracochlear membranes, hindering accurate segmentation.

Purpose of the Study:

• To develop an advanced segmentation technique for intracochlear cavities.
• To improve surgical planning for cochlear implant insertion.
• To enhance the likelihood of successful implant placement and hearing restoration.

Main Methods:

• An active shape model was created using ex vivo micro CT (μCT) scans of the cochlea.
• The model was applied to segment conventional CT scans, estimating unseen structures.
• The method leverages partial information from conventional scans to infer cavity boundaries.

Main Results:

• The developed method achieved an average Dice similarity coefficient of 0.75.
• Mean and maximum surface errors were reported as 0.21 mm and 0.80 mm, respectively.
• Quantitative evaluation demonstrated the efficacy of the active shape model approach.

Conclusions:

• The active shape model effectively segments intracochlear cavities, overcoming limitations of conventional imaging.
• This technique can aid surgeons in optimizing cochlear implant placement for better hearing outcomes.
• The method shows promise for improving surgical precision in treating profound hearing loss.