Enhancing nasal endoscopy: Classification, detection, and segmentation of anatomic landmarks using a convolutional...

Area of Science:

• Otorhinolaryngology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Nasal endoscopy (NE) is crucial for evaluating sinonasal diseases.
• Accurate identification of anatomical structures like turbinates is essential for diagnosis.
• Manual segmentation of turbinates in NE images is time-consuming and subjective.

Purpose of the Study:

• To develop and validate a deep learning model for automated turbinate localization and segmentation.
• To assess the accuracy of the proposed model in clinical NE images.
• To establish a foundation for AI-driven interpretation of endoscopic nasal findings.

Main Methods:

• A convolutional neural network (CNN) architecture was employed for image analysis.
• The model was trained on a dataset of nasal endoscopy images.
• Performance was evaluated using standard segmentation metrics.

Main Results:

• The CNN-based model achieved high accuracy in localizing and segmenting nasal turbinates.
• Quantitative analysis demonstrated the model's reliability in diverse NE images.
• The model showed potential for clinical application in turbinate analysis.

Conclusions:

• A CNN model provides an accurate and automated method for turbinate segmentation in NE.
• This AI approach can significantly enhance the efficiency and objectivity of NE interpretation.
• The developed model serves as a basis for advanced algorithms in endoscopic diagnostics.