Facial nerve image enhancement from CBCT using supervised learning technique.

Area of Science:

• Medical Imaging
• Neurosurgery
• Machine Learning

Background:

• Facial nerve segmentation is crucial for cochlear implantation surgery.
• Cone-beam CT (CBCT) is used for surgical planning but has low resolution, hindering facial nerve identification.
• Enhancing facial nerve visualization in CBCT is needed for improved surgical outcomes.

Purpose of the Study:

• To develop a supervised learning method to enhance facial nerve image information from CBCT scans.
• To improve the quality of facial nerve visualization in CBCT for better surgical planning.
• To facilitate accurate facial nerve segmentation using enhanced CBCT data.

Main Methods:

• A supervised learning approach utilizing a multi-output random forest model was employed.
• The model learned the mapping between lower-resolution CBCT images and higher-resolution micro-CT images.
• Evaluation involved qualitative and quantitative assessments using dedicated segmentation software (OtoPlan).

Main Results:

• The proposed approach demonstrated potential in enhancing facial nerve image quality from CBCT.
• Improved image quality facilitated better facial nerve segmentation.
• The enhanced images were successfully utilized with surgical planning software (OtoPlan).

Conclusions:

• Supervised learning can effectively enhance facial nerve image quality in CBCT.
• Improved CBCT imaging of the facial nerve can aid cochlear implantation surgical planning.
• This technique shows promise for leveraging facial nerve segmentation in clinical practice.