To develop and validate a deep learning model for screening adenoid hypertrophy (AH) using 2D facial photographs.
Approach:
Model Development: Deep learning models were trained and evaluated using a feature-fusion architecture, multi-view image analysis, and class-imbalanced dynamic sampling.
Key Findings:
The study utilized the largest dedicated facial dataset for AH screening to date.
A deep learning model was developed that analyzes facial features associated with AH.
Interpretation:
Limitations:
The study's retrospective design may introduce biases.
The model's performance is dependent on the quality and diversity of the dataset.
Conclusion:
The developed deep learning model provides a framework for non-invasive screening of adenoid hypertrophy in children.
