Diagnosis of Obstructive Sleep Apnea Using Dynamic Strain Measurements of Tongue Muscle
Date Issued
2020-09
Author(s)
Shih-Jou, Chung, Yu-Wen Huang, Pei-Lin Lee, and Pai Chi, Li
Abstract
Background, Motivation, Hypothesis/Goal, and Objectives
Tongue hypertrophy and upper airway obstruction are important factors causing obstructive sleep apnea (OSA).
However, direct ultrasound examinations of the retropharyngeal space have been challenging with limited clinical
performance. Currently, the apnea-hypopnea index (AHI) and polysomnography (PSG) are commonly used in the
clinical diagnosis but these procedures suffer from limited accuracy and/or long examination time. In this study, we
propose a new ultrasound functional imaging approach by measuring dynamics of the tongue thickness deformation
using 2D speckle tracking.
Statement of Contribution/Methods
The strains of the tongue muscle are measured as a function of time at pre-specified reference lines in both sagittal
and coronal planes. 2D speckle tracking is performed on submental ultrasound images when the patient is awake. 30
severe and 60 mild-moderate OSA patients, 8 health and 12 health volunteers underwent ultrasound measurements
of the tongue motion with normal breathing, Müller Maneuver and Maximum Tongue Pressure Measurement (TP),
as well as clinical and polysomnographic assessments at the National Taiwan University Hospital. MRI was also
performed for healthy volunteers. Two-third of patients were randomly selected for model-development group which
were used to construct model for predicting severe OSA. The results were validated in remaining one-third of patients.
Note that MM simulates the upper airway obstruction situation when awake and TP quantifies the Genioglossus
muscle collapse level. Speckle tracking is used to determine the extent of muscle contraction.
Results, Discussion and Conclusions
Speckle tracking was performed to find the dynamic strain curves at anterior tongue (A), posterior tongue base (P)
and maximum tongue base apex (TBA) during MM and TP as illustrated in Fig1. The TBA of TP results in Table 1
show that the proposed method is able to distinguish severe and non-severe OSA patients. The area under the receiver
operating characteristic curve is 0.81, and sensitivity is 73%, specificity is 78 % for mild-moderate patients, and
sensitivity is 80%, specificity is 81 % for severe patients (Table 2). The clinical potential of the proposed method for
OSA diagnosis is demonstrated.
Event(s)
2020 IEEE International Ultrasonics Symposium
Type
conference paper