Researchers have developed deep learning models, specifically the Audio Spectrogram Transformer (AST) and ResNet-50, to identify multi-level upper airway collapses in obstructive sleep apnea (OSA) patients using snoring sounds. The AST model demonstrated superior performance in detecting obstructions at the Velum, Oropharynx, and Retropalatal levels, achieving high scores in F1-score, Matthews Correlation Coefficient, and AUC. While performance was limited at other levels due to small sample sizes, the study suggests these acoustic profiling models offer a promising non-invasive tool for tracking airway dynamics and aiding clinical treatment planning. AI
IMPACT This research demonstrates a novel application of AI in non-invasively diagnosing medical conditions, potentially improving patient care and treatment planning.
RANK_REASON Research paper detailing a novel application of deep learning models for medical diagnosis. [lever_c_demoted from research: ic=1 ai=1.0]
- Audio Spectrogram Transformer
- Epiglottis
- obstructive sleep apnea
- Oropharynx
- ResNet-50
- Shang Ran Huang
- Synthetic Minority Over-sampling Technique
- Velum
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
