Facial Contour Features Measured on CT Reflects Upper Airway Morphology in Patients With OSA.

OBJECTIVE: To determine facial contour features, measured on computed tomography (CT), related to upper airway morphology in patients with obstructive sleep apnea (OSA); certain phenotype of facial abnormalities implying restriction of craniofacial skeleton and adipose tissue nimiety has predicted the value of the severity of OSA. MATERIALS AND METHOD: Sixty-four male patients with OSA [apnea-hypopnea index (AHI) ≥10/h] who had upper airway CT were randomly selected to quantitatively measure indicators of facial contour and upper airway structures. Pearson correlation analyses were performed. Partial correlation procedure was used to examine correlations while controlling body mass index (BMI). RESULTS: Upper airway anatomy can nearly all be reflected in the face, except retroglossal airway. Upper face width can be measured to assess the overall skeletal structures of the airway. Lower face width can be used to represent how much adipose tissue deposited. Hard palate, retropalatal, and hypopharyngeal airways have corresponding face indicators respectively. Midface width is a better predictor of AHI severity and minimum blood oxygen even than neck circumference because it contains the most anatomical information about the airway, including RP airway condition, soft palate length, tongue volume, etc. These correlations persisted even after correction for BMI. CONCLUSIONS: All anatomical features of the upper airway except retroglossal airway can be reflected in the face, and midface width is the best predictor of AHI severity and minimum blood oxygen, even better than neck circumference and BMI.

Prediction of obstructive sleep apnea using deep learning in 3D craniofacial reconstruction.

Background: Obstructive sleep apnea (OSA) is a common sleep disorder. However, current diagnostic methods are labor-intensive and require professionally trained personnel. We aimed to develop a deep learning model using upper airway computed tomography (CT) to predict OSA and to warn the medical technician if a patient has OSA while the patient is undergoing any head and neck CT scan, even for other diseases. Methods: A total of 219 patients with OSA [apnea-hypopnea index (AHI) ≥10/h] and 81 controls (AHI <10/h) were enrolled. We reconstructed each patient's CT into 3 types (skeletal structures, external skin structures, and airway structures) and captured reconstructed models in 6 directions (front, back, top, bottom, left profile, and right profile). The 6 images from each patient were imported into the ResNet-18 network to extract features and output the probability of OSA using two fusion methods: Add and Concat. Five-fold cross-validation was used to reduce bias. Finally, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were calculated. Results: All 18 views with Add as the feature fusion performed better than did the other reconstruction and fusion methods. This gave the best performance for this prediction method with an AUC of 0.882. Conclusions: We present a model for predicting OSA using upper airway CT and deep learning. The model has satisfactory performance and enables CT to accurately identify patients with moderate to severe OSA.

Classification of facial phenotypes in Asian patients with obstructive sleep apnea.

BACKGROUD: The facial phenotypes of Asian obstructive sleep apnea (OSA) patients remain unclear. OBJECTIVES: (1) To describe the facial features of OSA patients. (2) To develop a model based on facial contour indicators to predict OSA. (3) To classify the facial phenotypes of Asian OSA patients. MATERIALS AND METHODS: 110 patients with OSA (apnea-hypopnea index [AHI] ≥ 10/h) and 50 controls (AHI< 10/h) were selected to measure facial contour indicators. Indicators were compared between OSA patients and the control group. We used multivariable linear regression analysis to predict OSA severity and K-means cluster analysis to classify OSA patients into different phenotypes. RESULTS: We built a model to predict OSA which explained 49.1% of its variance and classified OSA patients into four categories. Cluster 1 (Skeletal type) had the narrowest facial width indicators with narrowing of the retroglossal airway. Cluster 2 (Obese type) had the widest face, and narrowest hard palate, retropalatal, and hypopharyngeal airways. Cluster 3 (Nose type) had the narrowest nasal cavity. Cluster 4 (Long type) had the longest airway length. CONCLUSIONS AND SIGNIFICANCE: Patients with OSA were classified into four categories, each of which identified different anatomic risk factors that can be used to select the treatment.