Automatic Detection of the Inner Ears in Head CT Images Using Deep Convolutional Neural Networks

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Otolaryngology

Background:

• Cochlear implants (CIs) restore hearing but require lengthy programming.
• Current programming relies on subjective patient feedback and lacks electrode position data.
• Image-guided cochlear implant programming (IGCIP) improves outcomes but needs automation.

Purpose of the Study:

• To develop an automated method for detecting ear presence in CT scans.
• To address variability in CT image acquisition for clinical deployment of IGCIP.
• To enable fully automated image processing for cochlear implant programming.

Main Methods:

• A deep learning approach was used for automated classification of head CT volumes.
• The model was trained to detect the presence of two ears, one ear, or no ear.
• The dataset comprised over 2,000 CT volumes from 153 patients.

Main Results:

• The deep learning model achieved 95.97% overall classification accuracy.
• The automated detection successfully identified ear presence in diverse CT volumes.
• This automation is crucial for the clinical deployment of IGCIP.

Conclusions:

• Deep learning offers a robust solution for automatically identifying relevant anatomy in CT scans.
• This automated detection facilitates the clinical application of image-guided cochlear implant programming.
• The developed method enhances the efficiency and accuracy of preparing CI patients for surgery.