ELHnet: a convolutional neural network for classifying cochlear endolymphatic hydrops imaged with optical coherence...

Area of Science:

• Otolaryngology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Endolymphatic hydrops is a key diagnostic marker for Meniere's disease.
• Non-invasive detection using optical coherence tomography (OCT) shows promise in animal models and clinical settings.
• Accurate classification of endolymphatic hydrops is crucial for effective Meniere's disease management.

Purpose of the Study:

• To develop and evaluate ELHnet, a deep convolutional neural network (CNN) for classifying endolymphatic hydrops.
• To utilize OCT imaging features for automated detection of endolymphatic hydrops in a mouse model.
• To establish a novel, AI-driven approach for endolymphatic hydrops classification.

Main Methods:

• Development of ELHnet, a CNN, trained on 2159 OCT images from 17 mice.
• Input data comprised image pixels and observer-determined labels of endolymphatic hydrops.
• Validation of ELHnet performance on an independent dataset of 37 mice.

Main Results:

• ELHnet achieved high classification accuracy, correctly identifying endolymphatic hydrops in 34 out of 37 mice.
• Demonstrated improved performance compared to previous computer-aided classification methods.
• This represents the first deep CNN specifically designed for endolymphatic hydrops classification.

Conclusions:

• ELHnet effectively classifies endolymphatic hydrops using OCT images in a mouse model.
• The developed AI tool shows significant potential for improving Meniere's disease diagnosis.
• This study highlights the advancement of deep learning in otological research and diagnostics.