Classification and Segmentation of Mucus Morphology Using Deep Learning During Diagnostic Nasal Endoscopy

Area of Science:

• Otolaryngology
• Medical Artificial Intelligence
• Computational Pathology

Background:

• Nasal endoscopy is crucial for diagnosing sinonasal conditions.
• Objective assessment of mucus in sinonasal disease is lacking.
• Current mucus evaluation relies on subjective interpretation.

Purpose of the Study:

• To develop and validate a deep learning model for mucus morphology classification.
• To assess the model's performance in segmenting and classifying different mucus types.
• To explore the utility of AI-driven mucus analysis for sinonasal inflammation detection.

Main Methods:

• Utilized a pretrained deep learning model for image segmentation and classification.
• Applied the model to mucus samples obtained during nasal endoscopy.
• Evaluated model performance across various mucus morphologies.

Main Results:

• The deep learning model demonstrated strong performance in segmenting and classifying mucus morphologies.
• Objective classification of mucus types was achieved.
• The model's findings correlate with the presence of sinonasal inflammation.

Conclusions:

• AI-powered objective classification of mucus morphology is feasible.
• This approach offers a novel tool for sinonasal inflammation detection.
• This study pioneers objective mucus morphology classification in nasal settings.