講師資料
Talks：
Rethinking Airway Geography: Decoding the Multifactorial Nature of OSA with Machine Learning
透過機器學習解碼阻塞型睡眠呼吸中止症上呼吸道的多因素特徵
Name：
陳裕豐醫師（高醫附醫）
Position：
Assistant professor
Affiliation：
1. Division of Oral and Maxillofacial Surgery, Department of Dentistry, Kaohsiung Medical University, Taiwan.
2. Sleep center, Kaohsiung Medical University, Taiwan.
Email：
Photo：
Research Interests：
Obstructive sleep apnea, Sleep surgery, Craniomaxillofacial surgery
Selected Publications：
1. Medication related osteonecrosis of the jaw. Yu-Feng Chen; Hong-Po Chang. New England Journal of Medicine. May 25, 2023; 388:e69. DOI: 10.1056/NEJMicm2209172.
2. Optimizing Mandibular Sagittal Split of Large Maxillomandibular Advancements for Obstructive Sleep Apnea: Patient and Surgical Factors. Yu-Feng Chen; Edward Chengchun Ko; Soroush Zaghi; Audrey Yoon; Ryan Williams; Robert Riley; Stanley Yung Chuan Liu. Clinical Oral Investigations. March 2020; 24(3):1359–1367. DOI: 10.1007/s00784-019-03017-5.
3. Obstructive sleep apnea treatment in adults. Yu-Feng Chen; Je-Kang Du; Hong-Po Chang. Kaohsiung Journal of Medical Sciences. 2020; 36(1):7–12.
4. Pediatric obstructive sleep apnea: computational fluid dynamics analysis of upper airway. Yu-Feng Chen; Ming-Hsuan Sheen; Hong-Po Chang; Yu-Chuan Tseng. Journal of Dental Sciences. June 2021; 17(1):589–591.
5. Hypoglossal nerve stimulation for treatment of obstructive sleep apnea (OSA): a primer for oral and maxillofacial surgeons. Sung Ok Hong; Yu-Feng Chen; Junho Jung; Yong-Dae Kwon; Stanley Yung Chuan Liu. Maxillofacial Plastic and Reconstructive Surgery. December 2017; 39:27.
6. The upper airway nasal complex: Structural contribution to persistent nasal obstruction. Ryan Williams; Vishal Patel; Yu-Feng Chen; Navarat Tangbumrungtham; Andrew Thamboo; Sam P. Most; Jayakar V. Nayak; Stanley Y. C. Liu. Otolaryngology–Head and Neck Surgery. March 2019; 161(1). DOI: 10.1177/0194599819838262.
7. A Novel Instrument-based Technique to Identify the Mandibular Foramen during Sagittal Split Ramus Osteotomy. Hsin-Yi Tsou; Yiu Yan Leung; Ying-An Chen; Yu-Ray Chen; Yu-Feng Chen. Plastic and Reconstructive Surgery – Global Open. August 2026; 14:e8061. DOI: 10.1097/GOX.0000000000008061.
Abstract：
Obstructive sleep apnea (OSA) presents a complex biomechanical challenge and is traditionally evaluated using polysomnography (PSG) for disease severity, drug-induced sleep endoscopy (DISE) for dynamic airway collapse, and cone-beam computed tomography (CBCT) for static anatomy. Previous studies examining the relationship between airway morphology and OSA severity have yielded inconsistent findings, partly because conventional linear regression may not adequately capture the nonlinear and multifactorial nature of the disease.
To address this limitation, a recent study applied a Random Forest machine-learning model to a cohort of 40 patients to evaluate the relative predictive importance of airway parameters for the apnea–hypopnea index (AHI). DISE identified complete oropharyngeal collapse as being associated with greater disease severity; however, patterns of dynamic obstruction at the velum, tongue base, and epiglottis showed limited correlation with overall AHI. Likewise, individual static CBCT parameters demonstrated limited predictive value when assessed using conventional linear regression. In contrast, nonlinear machine learning substantially improved prediction and highlighted maximal cross-sectional area (Max-CSA) as one of the most important predictors of OSA severity.
These findings offer a different perspective on airway morphology. Max-CSA measurements at the velum, oropharynx, and tongue-base levels showed greater predictive importance than either the narrowest airway dimension (Min-CSA) or most dynamic DISE parameters. This observation introduces the concept of structural reserve capacity: a larger maximal airway dimension may provide greater anatomical reserve, helping preserve airway patency as neuromuscular tone decreases during sleep. From a preoperative surgical perspective, CBCT may therefore provide information beyond static anatomical description, potentially serving as a useful tool for estimating overall disease severity and underlying airway resilience when interpreted within a nonlinear, multivariable framework.
2026年會：