Related Experiment Video
Updated: Jun 24, 2026
15:04
Modeling and Simulations of Olfactory Drug Delivery with Passive and Active Controls of Nasally Inhaled Pharmaceutical Aerosols
Published on: May 20, 2016
AI-Enabled Mucus Segmentation in Nasal Endoscopy with State Space and Attention-Based Modeling
Dipesh Gyawali1, Jonathan Bidwell1, Elena Karras2
1Department of Otorhinolaryngology Ochsner Health New Orleans Louisiana USA.
Laryngoscope Investigative Otolaryngology
|June 15, 2026
Summary
SUM-MucusNet, an AI model, improves mucus segmentation in nasal endoscopy images, overcoming challenges like poor quality and glare. This aids sinusitis diagnosis by providing more accurate mucus identification.
Area of Science:
- Medical Computer Vision
- Artificial Intelligence in Medicine
- Otolaryngology Imaging
Background:
- Accurate mucus segmentation in nasal endoscopy is crucial for sinusitis diagnosis.
- Existing models face challenges with thin mucus, boundary delineation, and illumination artifacts.
Purpose of the Study:
- Introduce SUM-MucusNet, an AI model to enhance mucus identification in challenging nasal endoscopy images.
- Improve the accuracy and reliability of mucus segmentation for clinical diagnosis.
Main Methods:
- Trained SUM-MucusNet, a deep learning model, on 3486 annotated endoscopy images from 454 patients.
- Evaluated performance using Dice coefficient, IoU, precision, and recall against expert observations and existing models.
Main Results:
- SUM-MucusNet achieved a Dice coefficient of 0.71, a statistically significant improvement over prior models.
- Correctly segmented 89.3% of mucus in poor-quality images, with better boundary detection and reduced false positives from glare/overexposure.
- The model operates in real-time at 17.3 frames per second.
Conclusions:
- SUM-MucusNet enhances mucus segmentation in nasal endoscopy by addressing image quality, boundary, and illumination issues.
- Advances medical computer vision in an under-explored area, supporting quantitative mucus assessment.
- Facilitates standardized diagnosis and longitudinal disease monitoring in nasal endoscopy.
