Association of Backscattered Ultrasonographic Imaging of the Tongue With Severity of Obstructive Sleep Apnea in Adults

Key Points

Question

Using standardized backscattered ultrasonographic imaging (BUI) analysis, do upper airway tissue characteristics correlate with severity of obstructive sleep apnea (OSA)?

Results

In this prospective, single-center, diagnostic study of 89 adult patients, BUI analysis demonstrated a strong association between the severity of OSA and backscattered intensity of the posterior tongue.

Meaning

There remains a need of a noninvasive and cost-effective method to visualize upper airway tissue to optimize treatment of OSA, and quantitative ultrasonography, such as BUI, could be of clinical utility.

Abstract

Importance

Determining interventions to manage obstructive sleep apnea (OSA) depends on clinical examination, polysomnography (PSG) results, and imaging analysis. There remains the need of a noninvasive and cost-effective way to correlate relevant upper airway anatomy with severity of OSA to direct treatment and optimize outcome.

Objective

To determine whether backscattered ultrasonographic imaging (BUI) analysis of the tongue is associated with severity of OSA in adults.

Design, Setting, and Participants

In this prospective, single-center, diagnostic study of a consecutive series of patients (aged ≥18 years) at a sleep surgery clinic, the 89 included patients had a PSG within 3 years at the time of ultrasonography and BUI analysis between July 2020 and March 2022. Patients were excluded if body mass index had changed more than 10% since time of PSG. A standardized submental ultrasonographic scan with laser alignment was used with B-mode and BUI analysis applied to the tongue. The B-mode and BUI intensity were associated with the apnea-hypopnea index (AHI), a measure of severity of apnea from normal (no OSA) to severe OSA.

Exposures

Ultrasonography and PSG.

Main Outcomes and Measures

The main outcomes were BUI parameters and their association with AHI value.

Results

Eighty-nine patients were included between July 2020 and March 2022. A total of 70 (78.7%) male patients were included; and distribution by race and ethnicity was 46 (52%) White participants, 22 (25%) Asian participants, and 2 (2%) African American participants, and 19 (21%) others. Median (IQR) age was 37.0 (29.0-48.3) years; median (IQR) BMI was 25.3 (23.2-29.8); and median (IQR) AHI was 11.1 (5.6-23.1) events per hour. At the middle to posterior tongue region, the 4 OSA severity levels explained a significant portion of the BUI variance (η² = 0.153-0.236), and a significant difference in BUI values was found between the subgroups with AHI values of less than 15 (no OSA and mild OSA) and greater than or equal to 15 (moderate OSA and severe OSA) events per hour. The echo intensity showed no significant differences. The BUI values showed a positive association with AHI, with a Spearman correlation coefficient of up to 0.43. Higher BUI values remained associated with higher AHI after correction for the covariates of BMI and age.

Conclusions and Relevance

In this prospective diagnostic study, standardized BUI analysis of the tongue was associated with OSA severity. With the practicality of ultrasonography, this analysis is pivotal in connecting anatomy with physiology in treatment planning for patients with OSA.

This prospective diagnostic study assesses whether ultrasonography with backscattered imaging analysis of the tongue correlates with severity of obstructive sleep apnea in adults.

Introduction

Obstructive sleep apnea (OSA) is a major public health burden associated with negative effects on cardiovascular, metabolic, cognitive, and mental health. The condition of OSA accounts directly, or is associated with, increased medical costs. The cost is even more staggering when the effects on society, such as lost productivity, are considered.^1,2,3,4 The prevalence in the adult population is high and variable, with contributions from age, sex, and race and ethnicity. Globally, more than a billion individuals are diagnosed with OSA.⁵ In the US, OSA is estimated to affect 38% of the population or more than 25 million Americans.^6,7

Polysomnography (PSG) remains the reference standard diagnostic tool for OSA. Diagnosis and severity are classified using the apnea-hypopnea index (AHI) that ranges from normal (no OSA) to severe OSA, according to the American Academy of Sleep Medicine.⁸ The PSG procedure provides a wealth of physiological data about sleep, but it does not readily lead to interpretation of the anatomic cause of OSA. It is adequate for assessing the efficacy of positive airway pressure therapy, the medical first-line management. However, less than 34% of users are adherent to positive airway pressure therapy long-term. Selection of other interventions, including surgery, require both anatomical and physiological information.⁹ Objective methods to connect physiological data from PSG with upper airway anatomy are key to precise interventions and improved outcomes for OSA.¹⁰

Imaging is widely used for anatomic endotyping. However, radiographic imaging and computed tomography (CT) have not shown consistent correlation with OSA severity, except for select craniofacial measurements.^11,12,13,14 Radiography and CT scans do not image upper airway muscles, including the tongue, as well as magnetic resonance imaging (MRI). For OSA, MRI has been used to study elastography, tongue fat segmentation, and dynamic collapse during sleep. The MRI has demonstrated significant correlation with OSA severity and treatment outcome.^{12,15,16,17,18,19} Drug-induced sleep endoscopy is another way to explore airway dynamics and sites of collapse. However, it requires an operating room setting and is cost prohibitive to be done at a population level. Nevertheless, it has contributed to decision-making in sleep surgery with good results. For example, our group has shown that lateral pharyngeal wall collapse is an independent predictor for oxygen desaturation, and it can be resolved by maxillomandibular advancement.^20,21 Studies also show that hypoglossal nerve stimulation is not as effective in patients with complete concentric collapse of the velum, and is in fact an exclusion criteria.^22,23 For real-world applications, CT, MRI, and drug-induced sleep endoscopy are not practical for the outpatient and general practice settings. However, what we have learned from these modalities is important as we seek practical alternatives.

Recently, a study correlating echo intensity of tongue ultrasonography with MRI intensity in patients with OSA demonstrated a high degree of correlation.²⁴ Concurrently, there are promising studies using backscattered ultrasonographic imaging (BUI) for other clinical indications with similar needs for practicality on a population level. Examples include BUI use to characterize cervical remodeling during pregnancy, bone healing, myocardium after infarct, and liver steatosis.^{25,26,27,28,29}

Optimization of ultrasonography may be a cost-effective way to repeatedly examine upper airway tissue as part of the longitudinal care of OSA. In this study, the tongue of patients across a wide range of age, body mass index (BMI), and AHI was scanned via a standardized ultrasonography protocol, and analyzed with backscattered imaging, to obtain a predictive model of OSA severity.

Methods

This prospective diagnostic study was conducted at Stanford University Sleep Surgery Clinic with approval by the institutional review board (IRB: 53172). Patients visiting the clinic from July 2020 to March 2022 were included if they were 18 years or older and provided written consent to participate in the study. Inclusion requires an attended or unattended PSG within 3 years at the time of ultrasonography. Patients whose BMI changed more than 10% between the ultrasonography and PSG were excluded.

Ultrasonography Setup

An ultrasound scanner cleared by the US Food and Drug Administration (FDA), Terason uSmart3200T (K193510; Teratech Corporation) with a convex transducer (5C2A), was used to obtain radiofrequency ultrasound signals and B-mode images. Radiofrequency ultrasound signals were analyzed with an FDA-cleared software AmCAD-US (K162574; AmCad BioMed Corporation). Laser beams (FDA Establishment Registration & Device Listing No. 3015218501; AmCad BioMed Corporation) were used to align supine patient position to the sagittal plane, the Frankfort horizontal plane (FH plane) and a cross-sectional plane through the hyoid bone and the external acoustic meatus (HM plane) (Figure 1A). The ultrasound transducer is aligned with the laser projection of the HM plane at the submental region to obtain transverse, cross-sectional images. The transducer performed automated scans to avoid bias in manual scanning, with a 30° section of the upper airway including the tongue regions labeled A, B, C, and D in Figure 1B.

Figure 1. Ultrasonography Equipment and Upper Airway Diagram.

A, Standardized submental ultrasonographic scanner with laser alignment, and B, the 30-degree sector region of upper airway swept by the automatic scan. The region to be analyzed with ultrasonography with backscattered imaging is shaded with purple color. Abbreviations: H, horizontal; HM, hyoid–external meatus.

BUI and Nakagami Parameter Analysis

Backscattered ultrasound signals were analyzed from a 1.7 cm × 1.7 cm region of interest (ROI) at the dorso-posterior tongue (Figure 1B). The analysis is performed by taking the absolute value of the radiofrequency ultrasound signal’s Hilbert transformation to obtain echo amplitude (intensity) data, which was then log-compressed to form the B-mode ultrasonographic image, as shown in Figure 2A and C. Nakagami parameter backscattered statistics was used to analyze the distribution of echo intensity.^26,28 The parameter value is a statistic of the Nakagami distribution estimated for the echo intensity data within a window in the ROI. The window, sized 0.3 cm × 0.3 cm, starts from the corner of the ROI, slides laterally and then axially with an overlap rate of 95% to cover the entire ROI. The Nakagami parameter calculation was repeated for each sliding window to construct the BUI color map (Figures 2B and D).

Figure 2. Ultrasonographic Images.

B-mode (log-compressed echo intensity) images (A and C), and backscattered ultrasonographic imaging (BUI) color maps (B and D) composed of the Nakagami parameter values computed from sliding window. The black squares indicate the region of interest for echo intensity measurement and BUI analysis. Images were acquired and computed from the B regions of patients with mild obstructive sleep apnea (A and B) and moderate obstructive sleep apnea (C and D).

The Nakagami parameter value (referred to as BUI value hereafter), ranging from 0 to 1, reflects the changes in the shape of echo intensity distribution from a longer right tail (BUI value <1) to a Rayleigh distribution (BUI value = 1).³⁰ A BUI value higher than 1 indicates a post–Rayleigh distribution with a shorter right tail. The BUI value provides a strong correlation between backscattered statistics and tissue microstructure.²⁶

Statistical Analysis

Median BUI and echo intensity values from the ROI at the tongue were used for statistical analysis. Comparisons of the medians measured for OSA groups of different severity were estimated using the effect size of the Kruskal-Wallis test (η²) and median difference. Spearman rank correlation (ρ) was used to assess the association between BUI, echo intensity, and OSA severity. Spearman partial correlation (ρ^a) was adjusted for age and BMI. MedCalc statistical software, version 19.0.4 (MedCalc Software Ltd) and R statistical software, version 4.2.2 (R Project for Statistical Computing) were used for analyses.

Results

Of the 89 patients included in the study, the median (IQR) age was 37.0 (29.0-48.3) years; median (IQR) BMI (calculated as weight in kilograms divided by height in meters squared) was 25.3 (23.2-29.8); 70 (78.7%) male patients were included; and distribution by race and ethnicity was 46 (52%) White participants, 22 (25%) Asian participants, and 2 (2%) African American participants, and 19 (21%) others, including Hispanic, Pacific Islander, and multiracial individuals. The median (IQR) AHI was 11.1 (5.6-23.1) events per hour; 18 (20%) had no OSA, 35 (39%) had mild OSA, 20 (22%) had moderate OSA, and 16 (18%) had severe OSA (Table). Age (ρ = 0.36; 95% CI, 0.16-0.53) and BMI (ρ = 0.43; 95% CI, 0.24-0.59) were weakly and moderately associated with AHI, respectively. The BUI values at tongue region A (ρ = 0.43; 95% CI, 0.24-0.59) and B (ρ = 0.43; 95% CI, 0.24-0.59) were moderately associated with AHI. Echo intensity at all regions did not show significant association with AHI. Strongly significant differences in BUI value were found between AHI values less than 15 (no and mild OSA) and AHI values of at least 15 (moderate and severe OSA) events per hour at tongue region A (median difference, 0.08; 95% CI, 0.05-0.12), and B (median difference, 0.05; 95% CI, 0.02-0.08). When patients were grouped according to clinically defined OSA severity levels, echo intensity of the severe OSA group was significantly higher; however, the Kruskal-Wallis effect size was rather small, with η² = 0.04 (95% CI, −0.03 to 0.23) and 0.05 (95% CI, −0.02 to 0.27), respectively (Figure 3A and 3B, for tongue regions A and B). The 4 OSA severity levels appeared to explain significant portions of BUI variance (Figure 3C and 3D) at tongue regions A and B, with the Kruskal-Wallis effect size η² = 0.24 (95% CI, 0.11-0.41) and 0.15 (95% CI, 0.05-0.34), respectively. Two pairs of echo intensity and BUI color maps from patients with mild and moderate OSA are shown in Figure 2. Difference in the visual representation of BUI values is more prominent than that of the B-mode image (echo intensity map).

Table. Patient Obstructive Sleep Apnea Severity and Analyses.

Variable	Overall	OSA				η2 (95% CI)a	Spearman ρ (95% CI)b
		None	Mild	Moderate	Severe
Sample size, No.	89	18	35	20	16	NA	NA
Echo intensity, A region, median (IQR)	173.83 (140.23 to 210.70)	168.90 (132.36 to 197.40)	166.00 (141.37 to 212.75)	184.71 (139.34 to 200.90)	220.41 (162.61 to 280.79)	0.038 (−0.03 to 0.23)	0.20 (−0.01 to 0.39)
Echo intensity, B region, median (IQR)	220.41 (179.16 to 279.41)	206.43 (183.17 to 232.02)	222.57 (177.75 to 279.99)	198.70 (166.97 to 249.23)	319.42 (193.47 to 367.74)	0.047 (−0.02 to 0.27)	0.20 (−0.01 to 0.39)
Echo intensity, C region, median (IQR)	253.72 (205.93 to 309.70)	255.88 (234.41 to 319.31)	260.82 (209.54 to 306.91)	253.04 (177.33 to 296.86)	234.26 (203.81 to 321.18)	−0.026 (−0.03 to 0.10)	−0.01 (−0.22 to 0.20)
Echo intensity, D region, median (IQR)	217.44 (163.01 to 288.87)	226.04 (164.50 to 327.85)	212.34 (174.69 to 277.03)	235.83 (135.65 to 304.82)	234.11 (136.75 to 279.95)	−0.032 (−0.03 to 0.08)	−0.01 (−0.22 to 0.20)
BUI value, A region, median (IQR)	0.83 (0.76 to 0.88)	0.77 (0.70 to 0.84)	0.78 (0.73 to 0.86)	0.84 (0.80 to 0.87)	0.89 (0.86 to 0.92)	0.24 (0.11 to 0.41)	0.43 (0.24 to 0.59)
BUI value, B region, median (IQR)	0.83 (0.77 to 0.88)	0.77 (0.74 to 0.82)	0.82 (0.77 to 0.87)	0.87 (0.80 to 0.89)	0.84 (0.81 to 0.89)	0.15 (0.05 to 0.34)	0.43 (0.24 to 0.59)
BUI value, C region, median (IQR)	0.82 (0.76 to 0.86)	0.80 (0.74 to 0.82)	0.81 (0.76 to 0.86)	0.83 (0.77 to 0.86)	0.83 (0.79 to 0.87)	0.04 (−0.02 to 0.20)	0.25 (0.04 to 0.43)
BUI value, D region, median (IQR)	0.80 (0.75 to 0.84)	0.77 (0.76 to 0.84)	0.80 (0.75 to 0.84)	0.82 (0.78 to 0.85)	0.82 (0.75 to 0.87)	−0.01 (−0.03 to 0.12)	0.17 (−0.04 to 0.36)

Abbreviations: AHI, apnea-hypopnea index; BUI, backscattered ultrasonographic imaging; OSA, obstructive sleep apnea.

^a Effect size of Kruskal-Wallis test. η² = 0.01, 0.06, and 0.14 indicate a small, medium, and large effect size, respectively.

^b Association with AHI.

Figure 3. Echo Intensity, BUI, and AHI.

A and B, Echo intensity and apnea-hypopnea index (AHI). C and D, backscattered ultrasonographic imaging (BUI) measurements of the 4 severity groups for the A and B regions and AHI. The middle bar is the median with the top and lower sections of the boxes as the upper and lower quartiles. The whiskers indicate the 95% CIs, and the open circles indicate the outliers.

When AHI is not used to divide patients into clinically defined groups, but as a continuous variable, BUI values at tongue regions A and B showed positive association with AHI (ρ = 0.43; 95% CI, 0.24-0.59 for both regions). After correction of the covariates BMI and age, BUI values were still associated with AHI, with Spearman partial correlation ρ^a of 0.32 (95% CI, 0.12-0.50) and 0.39 (95% CI, 0.19-0.55) (Figure 4C and D). As compared with BUI, echo intensity was weakly associated with AHI after correction for the covariates at both regions with ρ^a = 0.05 (95% CI, −0.16 to 0.26) and 0.10 (95% CI, −0.12 to 0.30) (Figure 4A and B). Because the number of patients who underwent attended PSG (64) is substantially greater than those who underwent home sleep study (25), comparison between attended PSG and home sleep study across all clinical subgroups cannot be made with the current sample size.

Figure 4. Correlation Between Echo Intensity/BUI and AHI.

Regression plots and Spearman correlation statistics of the association between echo intensity (A and B), backscattered ultrasonographic imaging (BUI) measurements (C and D), and apnea-hypopnea index (AHI). Dotted curve indicates 95% CIs.

^aSpearman partial correlation after controlling for body mass index and age.

Discussion

In this prospective diagnostic study, we used BUI to analyze the upper airway tissue in patients with OSA. With a standardized ultrasonography protocol to limit operator bias, we demonstrated how tissue microstructure of the tongue can be measured to correlate with OSA severity. This was performed in a noninvasive, quantifiable, and repeatable fashion. Results of our early efforts show that even with just the dorsum of the tongue, BUI analysis of the microstructure can differentiate OSA severity. The predictive model of OSA severity is robust when BUI parameters are included with age and BMI. This is a pivotal first step in promoting the use of ultrasonography with novel analysis to study OSA pathogenesis.

Most in vivo ultrasonographic techniques to characterize human tissue are restricted to attenuation properties of the organ examined.³¹ There is greater agreement regarding the association of increased fat content, vs fibrosis, with ultrasonic attenuation coefficient. In contrast to this brightness (B-mode) grayscale ultrasonography that provides qualitative information, BUI is a type of quantitative ultrasonography. Specifically, BUI represents the information from ultrasound waves and tissue microstructure. The backscatter coefficient is the backscatter intensity returned by a tissue within a region of interest, expressed in cm/steradian.³² As BUI is related to the brightness on grayscale images, it is more sensitive to subtle changes in tissue pathology. Clinical research that has already taken advantage of BUI include noninvasive characterization of the following: (1) carotid arteries, (2) breast cancer cells,³³ (3) pediatric and adult hepatic fat,^34,35,36 (4) cancellous and cortical bone,^27,37,38 and (5) cervical remodeling during pregnancy.²⁵ The most mature use of BUI in clinical practice is for diagnosing and grading nonalcoholic fatty liver disease, which the American Institute of Ultrasound in Medicine and Radiological Society of North America Quantitative/Imaging Biomarkers Alliance are actively working on standardizing.³⁹

Previously, B-mode sonographic data were used to predict moderate to severe OSA using tongue area measurement during nasal breathing and Müller maneuver.⁴⁰ Normal controls and patients with mild OSA tend to demonstrate bidirectional tongue motions during a transition from normal breathing to the Müller maneuver. This finding is supported by MRI studies showing that patients with severe OSA exhibit minimal respiratory-related movement of the posterior tongue during wakefulness.⁴¹ This may highlight one of the pathophysiological changes in OSA, which is motor neuropathy in severe OSA. The difference between tongue base thickness has also been reported to be another independent predictor of OSA severity via ultrasonography.⁴² These studies show that B-mode ultrasonography can provide assessment of the retroglossal airway in OSA in an accessible and noninvasive manner. In the present study, while B-mode ultrasonographic imaging also identified severe OSA, BUI analysis enabled robust differentiation of mild, moderate, and severe OSA.

By enabling characterization of tissue associated with severity of OSA, upper airway BUI can be useful for directing and monitoring surgical procedures and their outcomes. When anatomic interventions are considered in OSA, the goal is to reduce collapsibility of the upper airway and improve airflow: essentially to change the negative pressure necessary to collapse the airway.^43,44 Other contributors to the pathogenesis of OSA such as arousal threshold, loop gain, and muscle tone are not mechanisms that surgery can address.⁴⁵ Because the focus is on the collapsibility of the upper airway, a way to visualize and characterize the upper airway muscles is critically important.^20,21 While ultrasonography can characterize collapsibility, there are many variables that affect its reproducibility. With BUI analysis of key dilator muscles, such as the tongue in this study, there is an objective measure of tissue microstructure that is associated with AHI.

Limitations

Limitations to the present study’s findings include small sample size, single-institution experience, underrepresentation of the female sex, and unequal racial and ethnic distribution. For now, it is only focused on the tongue. Additionally, a minority of participants underwent home sleep study, which tends to underestimate severity of OSA as compared with attended PSG. Nevertheless, to our knowledge, this is the most diverse population of adult patients with OSA that have been examined with ultrasound technology to date. Analysis of trends also show that home sleep study does underestimate BUI, hence in an expected direction. Future research with a larger study population, balanced proportions of sex and race, and examination of other upper airway muscles implicated in OSA pathogenesis will likely yield improvement in prognostic value using quantitative ultrasound technology. Subsequently, the potential role of using BUI to screen for surgical eligibility will be an exciting addition in the era of preservation pharyngoplasty, patient-specific maxillomandibular advancement, and hypoglossal nerve stimulation.^{46,47,48,49,50}

Conclusion

Standardized ultrasonography of the tongue with backscattered imaging analysis yields strong correlation with clinical severity of OSA. With the advantages associated with noninvasiveness and cost in the use of ultrasonography, this analysis is pivotal in reducing the gap between anatomy and physiology in clinical decision-making for OSA treatment.

Supplement 1.

Data Sharing Statement