Skip to main content
NCBI home page
Sleep logoLink to Sleep
. 2021 Dec 28;45(2):zsab295. doi: 10.1093/sleep/zsab295

A comparison of ultrasound echo intensity to magnetic resonance imaging as a metric for tongue fat evaluation

Jason L Yu 1,2, Andrew Wiemken 2, Susan M Schultz 3, Brendan T Keenan 2, Chandra M Sehgal 3, Richard J Schwab 2,✉
PMCID: PMC8842321  PMID: 34963001

Abstract

Study Objectives

Tongue fat is associated with obstructive sleep apnea (OSA). Magnetic resonance imaging (MRI) is the standard for quantifying tongue fat. Ultrasound echo intensity has been shown to correlate to the fat content in skeletal muscles but has yet to be studied in the tongue. The objective of this study is to evaluate the relationship between ultrasound echo intensity and tongue fat.

Methods

Ultrasound coronal cross-sections of ex-vivo cow tongues were recorded at baseline and following three 1 mL serial injections of fat into the tongue. In humans, adults with and without OSA had submental ultrasound coronal cross-sections of their posterior tongue. The average echo intensity of the tongues (cow/human) was calculated in ImageJ software. Head and neck MRIs were obtained on human subjects to quantify tongue fat volume. Echo intensity was compared to injected fat volume or MRI-derived tongue fat percentage.

Results

Echo intensity in cow tongues showed a positive correlation to injected fat volume (rho = 0.93, p < .001). In human subjects, echo intensity of the tongue base strongly correlated with MRI-calculated fat percentage for both the posterior tongue (rho = 0.95, p < .001) and entire tongue (rho = 0.62, p < .001). Larger tongue fat percentages (rho = 0.38, p = .001) and higher echo intensity (rho = 0.27, p = .024) were associated with more severe apnea-hypopnea index, adjusted for age, body mass index, sex, and race.

Conclusions

Ultrasound echo intensity is a viable surrogate measure for tongue fat volume and may provide a convenient modality to characterize tongue fat in OSA.

Keywords: upper airway, obstructive sleep apnea, Dixon MRI, echo intensity, ultrasound

Statement of Significance.

Tongue fat is a contributing factor to obstructive sleep apnea. Quantifying tongue fat may provide valuable information for the management of the disease. Magnetic resonance imaging (MRI) is the standard in quantifying tongue fat but is time-consuming and expensive. Our study shows that ultrasound echo intensity increases with increasing fat content in the tongue and strongly correlates with MRI-derived tongue fat measures. Ultrasound may be used as an alternative method to quantify tongue fat and is a modality that is faster, cheaper, and more easily accessible than MRI.

Introduction

Obstructive sleep apnea (OSA) is a condition that affects over 25 million Americans [1]. Risk factors for the development of OSA include obesity, age, and male gender [2]. The anatomy of the upper airway influences OSA risk, with studies showing that the size of soft tissue structures such as the tongue, soft palate, and lateral pharyngeal walls contribute to OSA pathogenesis [3, 4]. Magnetic resonance imaging (MRI) can provide an in-depth characterization of the tongue, including fat content, which has been shown to correlate with OSA severity [5]. Furthermore, among OSA patients who lose weight, a reduction in tongue fat is a key mediator of improvement in apnea-hypopnea index (AHI), suggesting that tongue fat reduction may be an important therapeutic target for treating OSA [6].

The use of MRI to characterize the tongue remains expensive and time-consuming. Ultrasound is a noninvasive, low risk, and inexpensive imaging modality that can also evaluate the tongue. Prior studies have used submental ultrasound to characterize the tongue in OSA, showing that increased tongue width and tongue thickness are associated with severe OSA [7–10]. In addition, ultrasound can capture dynamic motion that can characterize changes in tongue morphology with activities such as the Müller’s maneuver [11, 12].

One unexplored area in characterizing the tongue with ultrasound is the use of echo intensity. Echo intensity is a measure of the amount of sound wave reflected through tissues that can be used to evaluate skeletal muscle composition [13]. Studies have shown that increased echo intensity is associated with increases in fat related to atrophy or pathology in skeletal muscles [14–18]. Ultrasound echo intensity has also correlated to MRI-derived measurements of fat content in lower extremity skeletal muscles [19].

Previous studies have not used ultrasound echo intensity to characterize tongue fat. Ultrasound allows for a more accessible means of evaluating tongue fat, which may prove valuable in the diagnosis and management of OSA. The primary aim of this study was to determine if increased ultrasound echo intensity is a surrogate marker of increased fat content in the tongue in an ex-vivo cow tongue model and to verify this relationship among human subjects by comparing ultrasound echo intensity to MRI-derived tongue fat percentages. Our secondary aim was to show agreement between linear measurements of tongue width and depth obtained by ultrasound and MRI. We then explored the associations of tongue dimensions, ultrasound echo intensity, and tongue fat percentage with severity of the AHI. We hypothesize that increased tongue fat will lead to an increase in ultrasound echo intensity, that ultrasound echo intensity will correlate with MRI-derived tongue fat percentage, and that higher echo intensity would be associated with more severe AHI.

Methods

Cow tongue ultrasound

Confirmation that increasing fat deposition is associated with increased echo intensity was performed in five ex-vivo cow tongues. Baseline coronal cross-sectional recording of the tongues was captured on ultrasound by a single reviewer (J.L.Y.). Ultrasound recordings were obtained using an Ultrasonix Sonixtouch ultrasound (Ultrasonix Medical Corporation, BC, Canada). A curvilinear transducer (C5-2/60) was used for evaluation. The settings remained constant through all cow tongues to ensure comparisons of echo intensity were possible. Ultrasound imaging was performed in Harmonic (H) mode at 5 Megahertz (MHz) with a Gain setting of 60% and Depth of 6.0 cm. No adjustments to time gain compensation were made between specimens. Baseline coronal cross-sections of ex-vivo cow tongues were recorded, followed by repeated imaging after three serial injections of 1 milliliter (mL) of fat (Crisco) into the tongue, for a total of 3 mL of fat.

Patient selection

Subjects in this study were enrolled from a single institution as part of the Extreme Phenotypes in OSA (EXPO) study, a prospective study characterizing OSA through a variety of anatomic and physiologic metrics. The study was approved by the Institutional Review Board of the University of Pennsylvania (Study#823038). Subjects over the age of 18 years with (AHI ≥ 5 events/hour) and without (AHI < 5 events/hour) OSA were eligible for inclusion. Both obese and nonobese subjects with and without OSA were included in the study. Any subject unable to get an MRI (pacemaker, etc.), with a history of congestive heart failure, or with a recent cerebrovascular accident (<6 months) was ineligible for the study.

Submental ultrasound

Just prior to their scheduled MRI, subjects had submental ultrasound performed. A coronal cross-section of the posterior tongue was recorded using the same ultrasound and transducer described above. Settings to optimize imaging of the tongue on submental ultrasound were set with the help of a trained research ultrasonographer (S.M.S.). Recording of the ultrasound was performed by a single evaluator (J.L.Y.) who was blinded to the MRI results during the ultrasound acquisition. Settings of the ultrasound were not altered between patients to allow for comparisons of echo intensity. Ultrasound imaging was performed in Harmonic mode at 5 MHz with a Gain setting of 45% and a Depth of 9.0 cm. No adjustments to time gain compensation were made between participants. Patients were asked to sit and extend their neck with their mouth closed to facilitate probe placement under the chin while limiting interference from the mandible and hyoid. The probe was placed in a transverse direction to obtain a coronal section of the tongue (see Figure 1 for example of ultrasound probe placement and imaging). The probe was moved posteriorly along the neck resting just anterior to the hyoid bone and angled posteriorly to visualize the tongue base.

Figure 1.

Figure 1.

Open in a new tab

Placement of the probe and relevant anatomy visualized with submental ultrasound.

Ultrasound analysis

Analysis of ultrasound recordings and calculation of echo intensity was performed by a single reviewer (J.L.Y.) who had also performed the submental ultrasound recordings and was blinded to the results of the MRI analysis during echo intensity calculation. Video clips recordings were reviewed offline using Windows Media Player (Microsoft Corporation, Redmond, WA) and screenshots of a cross-section of the tongue base were captured. Optimal cross-sectional images were visually determined based on clarity of the borders around the tongue, maximum brightness of the tongue region, and minimum shadowing artifact. ImageJ software (U.S. National Institutes of Health, Bethesda, MD) was used to evaluate the width, height, and average echo intensity of the tongue cross-section. To calculate echo intensity, a circumferential outline of the tongue was drawn in ImageJ and the echo intensity was calculated as an average gray scale value of the pixels within the tongue outline reported as a unitless value (see Figure 2 for example of echo intensity calculation in a human tongue ultrasound).

Figure 2.

Figure 2.

Open in a new tab

Example of echo intensity calculation. (A) A line (red) was drawn bordering the edges of the tongue and the mean gray scale intensity (B) within the tongue was recorded.

MRI

MRI of the head and neck was obtained using a 1.5 Tesla MAGNETOM Espree Scanner (Siemens Medical Systems, Malvern, PA). Images were analyzed using Amira 4.1.2 (Visage Imaging, San Diego, CA). MRI of the upper airway was manually examined by a single reviewer (A.W.) who was blinded to the results of the ultrasound measurements. Tongue fat quantification was performed as described previously with Dixon imaging [6]. In short, tongue volumes were obtained from T1 spin-echo axial images. The resulting boundaries were then superimposed onto the axial Dixon images used to quantify tongue adipose tissue volume for the entire tongue, as well as specifically for the region of the tongue base, defined as all tissue proximal to the most posterior-inferior point of tongue, within one-third of the length of the long axis (see Figure 3 for the region of tongue base incorporated into the calculation). Tongue fat was measured on MRI by referencing the signal intensity of individual tongue voxels against the highest intensity voxels in the parapharyngeal fat pads, producing fat percentages for each tongue voxel that were compiled into a total adjusted fat volume. Height and width measurements were taken at a single coronal cross-section of the tongue at the start of posterior tongue base.

Figure 3.

Figure 3.

Open in a new tab

(A) Mid-sagittal MR slice showing region of interest. Region (shaded in pink) is defined as all tissue proximal to the most posterior-inferior point of tongue (red dot), within one-third of the length of the long axis. (B) Approximate position of ultrasound cross-section shown for reference.

Reproducibility analysis

Thirteen participants had repeat submental ultrasound, which occurred after completion of their MRI (approximately 1 h between ultrasounds), and three participants had repeat submental ultrasound performed at >1 month during a validation study. The ultrasound analysis was also repeated within a separate group of patients (n = 60) on the same recordings >1 month after initial analysis to demonstrate reliability of the analysis technique. Reliability/reproducibility was assessed using intraclass correlation coefficients (ICCs) calculated from the repeated echo intensity measurements. Following guidelines suggested by Landis and Koch [20], ICC values can be used to assess reliability/reproducibility as poor (<0.00), slight (0.00–0.20), fair (0.21–0.40), moderate (0.41–0.60), substantial (0.61–0.80), or almost perfect (0.81–1.00). Reproducibility of the MRI tongue fat analysis has been previously validated [6].

Statistical analysis

To evaluate the relationship between ultrasound echo intensity and injected fat volume in the cow tongue model, we applied repeated measures correlation analysis using the R/rmcorr package [21]. This approach accounts for the nonindependence of repeated measurements obtained within each cow tongue, providing an average correlation across samples while allowing subject-specific intercepts [21]. Among human subjects, continuous variables were summarized with means and SDs, and categorical variables using frequencies and percentages. Clinical data were compared between patients with and without OSA using t-tests (for continuous data) or chi-squared tests (for categorical data). Unadjusted Pearson’s correlations were used to evaluate the relationship between echo intensity and MRI-derived tongue fat percentage (posterior and whole tongue). Bland-Altman analysis was used to compare agreement between tongue width and height measurements obtained by ultrasound and MRI, including calculations of the mean difference (95% confidence interval [CI]), Bland-Altman limits of agreement (equal to ±2 SDs of the mean difference), and the correlation between difference and mean of the two methods [22]. Associations of ultrasound and MRI measurements with severity of AHI were evaluated using unadjusted Pearson’s correlations and partial correlations adjusted for age, sex, body mass index (BMI), and race (white vs. non-white). As echo intensity was considered a primary measure of interest, statistical significance was based on a p < .05. For secondary measures from ultrasound (tongue height, width, area) or MRI measurements (tongue fat percentages [posterior and whole tongue], height width, and area), statistical significance was determined using the Hochberg step-down method [23] with overall type I error maintained at 0.05.

A total of 83 participants (n = 28 without OSA [AHI < 5 events/hour] and 55 with OSA [AHI ≥ 5]) completed ultrasound measurements. MRI data on tongue fat percentage was unable to be obtained for n = 9 participants, resulting in 74 total participants with ultrasound and MRI (n = 24 without OSA and 50 with OSA). Given our primary focus on comparing ultrasound echo intensity and MRI tongue fat percentage, primary analyses were performed in the cohort with both measurements. Given this, we had at least 80% power to detect a moderate correlation of 0.32, as defined by Cohen [24], at an α = 0.05. Thus, our sample size resulted in good power for meaningful effect sizes, but nonsignificant results for smaller associations should be interpreted with some caution.

Results

Cow tongue analysis

Ultrasound echo intensity measurements with increasing injected fat volume within five ex-vivo cow tongues are illustrated in Figure 4. Repeated measures correlation analysis demonstrated a high correlation between increasing echo intensity values and increasing injected fat volume (rho [95% CI] = 0.93 [0.80 to 0.98], p < .001; Figure 4).

Figure 4.

Figure 4.

Open in a new tab

(A) Tongue ultrasound images at baseline and then after 1 cc (cubic centimeter), 2 cc, and 3 cc of fat (Crisco) were injected. (B) Measures of ultrasound echo intensity with increasing fat injections across five cow tongues. Data from ultrasound measurement are shown within five cow tongues, with repeated measurements at baseline and after 1cc, 2cc, and 3cc of injected fat. Analyses show a strong relationship between ultrasound and injected fat volume, with a repeated measures correlation (95% CI) of 0.932 (0.795 to 0.979).

Human analysis sample

Between October 31, 2017 and April 6, 2021, a total of 142 subjects were enrolled in the EXPO study. Of these, 83 (58.5%) had submental ultrasound. Three of these subjects were unable to tolerate the MRI portion of the study and six subjects had significant artifacts on the MRI that prevented tongue fat analysis. Tongue fat percentage was quantifiable in 74 (89.2%) participants. As shown in Table 1, participants were on average middle-aged (49.8 ± 12.4 years), obese (BMI of 31.6 ± 6.5 kg/m2), and a majority male (60.2%) and White (62.7%). Patients with OSA (AHI ≥ 5 event/hour) were older (p = .009) and more likely to be male (p = .0002). The BMI range of all participants was between 20.3 kg/m2 and 44.0 kg/m2, although the OSA and non-OSA groups had similar average BMI (p = .645). There were no differences in the racial distributions (p = .725). Differences in tongue width measurements and whole tongue fat percentage, but not echo intensity, posterior tongue fat percentage, or tongue height were observed between patients with and without OSA (see Table 1).

Table 1.

Comparison of characteristics in apneics and non-apneics

Characteristic Overall Non-OSA
(AHI < 5)		 OSA
(AHI ≥ 5)		 P
Sample Size, N 83 28 55 —
Age, years 49.8 ± 12.4 44.9 ± 11.9 52.4 ± 11.9 .009
BMI, kg/m2 31.6 ± 6.5 32.0 ± 6.2 31.3 ± 6.7 .645
Male, n (%) 50 (60.2) 9 (32.1) 41 (74.6) .0002
Race, n (%) .725
 White 52 (62.7) 16 (57.1) 36 (65.5)
 Black/African American 25 (30.1) 10 (35.7) 15 (27.3)
 Other 6 (7.2) 2 (7.1) 4 (7.3)
AHI, events/h 23.1 ± 24. 5 2.0 ± 1.3 33.9 ± 23.7 <.0001
Study type, n (%) .478
 HSAT 37 (44.6) 14 (50.0) 23 (41.8)
 PSG 46 (55.4) 14 (50.0) 32 (58.2)
Ultrasound measures
 Echo intensity 72.36± 11.22 71.62 ± 13.42 72.74 ± 10.04 .697
 Tongue height, mm 33.11 ± 5.22 32.45 ± 5.24 33.45 ± 5.22 .417
 Tongue width, mm 40.79 ± 6.19 38.80 ± 4.92 41.80 ± 6.55 .022
MRI measures
 Posterior tongue fat*, % 0.38 ± 0.06 0.36 ± 0.06 0.39 ± 0.06 .103
 Whole tongue fat*, % 0.30 ± 0.07 0.27 ± 0.06 0.31 ± 0.07 .005
 Tongue height†, mm 33.17 ± 5.40 32.12 ± 4.62 33.72 ± 5.73 .211
 Tongue width†, mm 41.03 ± 6.56 39.00 ± 4.77 42.09 ± 7.14 .035
Open in a new tab

HSAT = home sleep apnea testing; PSG = polysomnogram.

*n = 24 non-OSA, n = 50 OSA with available data.

† n = 24 non-OSA, n = 46 OSA with available data

Reproducibility analysis in humans

Sixteen subjects had repeated submental ultrasounds performed by a single evaluator (J.L.Y.) to evaluate reproducibility. Analyses resulted in an ICC (95% CI) of 0.978 (0.940 to 0.992), demonstrating almost perfect agreement with repeated ultrasound. In addition, a total of 60 participants had repeated echo intensity analysis of the same ultrasound scan. Data showed an ICC (95% CI) of 0.973 (0.955 to 0.984), supporting near-perfect reliability of the analysis technique.

Correlation between MRI tongue fat percentage and ultrasound echo intensity

Tongue fat analysis was unable to be calculated on 9 (10.8%) subjects because of subject intolerance or significant artifact on MRI. Among the 74 subjects with available tongue fat and ultrasound measurements, echo intensity of the tongue base obtained by ultrasound showed a strong positive correlation with calculations of fat percentage in the posterior tongue (rho = 0.95, p < .001) and the whole tongue (rho = 0.62, p < .001) from MRI (see Figure 5). The percentage of fat in the posterior tongue based on MRI showed a similar correlation with percentage fat in the whole tongue (rho = 0.69, p < .001) as was observed between ultrasound echo intensity and whole tongue fat percentage.

Figure 5.

Figure 5.

Open in a new tab

Ultrasound echo intensity showed a strong positive correlation to calculations of fat percentage obtained from MRI in both (A) the posterior tongue (rho = 0.95, p < .0001) and (B) the tongue as a whole (rho = 0.62, p < .0001). Posterior tongue fat percentage demonstrated a similar magnitude of correlation with whole tongue percentage (rho = 0.69, p < .0001) as was observed for ultrasound echo intensity.

Agreement between MRI and ultrasound tongue size measurements

There was good agreement between tongue width and height measurements quantified using MRI or ultrasound based on Bland-Altman analysis (see Figure 6). For tongue width (Figure 6A), there was a nonsignificant mean difference of −0.17 mm (95% CI: −0.45 to 0.11; p = .231) between ultrasound and MRI. Bland-Altman limits of agreement were equal to −2.53 to 2.19 mm, suggesting 95% of measurements taken with both techniques fell within 2.5 mm of one another. There was a moderate correlation between the difference and overall average between the two techniques (rho = −0.31, p = .010), suggesting the difference becomes more negative (e.g. larger underestimation by ultrasound) for larger values of tongue width. For tongue height (Figure 6B), there was no significant difference between the techniques (mean difference [95% CI] = 0.13 [−0.10 to 0.36]; p = .256) with limits of agreement ranging from −1.80 to 2.07 mm. There was no evidence of an association between the mean difference and magnitude of tongue height (rho = −0.11, p = .346). To understand the potential importance of these differences, we estimated the standardized mean differences represented by these limits of agreement based on the overall SD observed in our sample. As shown in Table 1, we observed a SD of 6.56 mm in tongue width and of 5.40 mm in tongue height. Thus, our observed limits of agreement for tongue width of −2.53 to 2.19 encompass standardized mean differences of −0.386 to 0.334. Similarly, for tongue height, the limits of agreement of −1.80 to 2.07 represent standardized mean differences of −0.333 to 0.383. Using guidelines provided by Cohen for small (0.2), medium (0.5), and large (0.8) standardized mean differences [24], both ranges represent small to medium differences.

Figure 6.

Figure 6.

Open in a new tab

Results of Bland-Altman analysis evaluating agreement between tongue width and height measurements from ultrasound and MRI. (A) Comparison of tongue width measurements show a nonsignificant mean difference of −0.17 mm (p = .231) with Bland-Altman limits of agreement ranging from −2.53 to 2.19 mm, suggesting approximately 95% of values are expected to be within 2.5 mm using either technique. There was a moderate correlation of −0.31 (p = .010) between the mean and the difference, suggesting the difference becomes more negative (e.g. larger underestimation by ultrasound) for larger values of tongue width. (B) Comparison of tongue height measurements show a nonsignificant difference between ultrasound and MRI, with Bland-Altman limits of agreement equal to −1.80 to 2.07 mm, suggesting approximately 95% of measurements are expected to be within 2 mm using either technique. There was no evidence of a significant relationship between bias and average values for tongue height.

Association of tongue measurements with AHI severity

Associations of tongue measurements with AHI severity were evaluated within the sample with both ultrasound and MRI to facilitate comparisons, using both unadjusted Pearson’s correlations and partial correlations controlling for differences in age, sex, BMI, and race (Table 2). Despite nonsignificant differences based on an AHI cut-point of 5 events/hour (see Table 1), in unadjusted analyses of OSA severity higher values of all tongue measurements except tongue height were significantly correlated with more severe AHI. After controlling for covariates, larger whole tongue (rho = 0.38, p = .001) and posterior tongue (rho = 0.28, p = .020) fat percentages and higher echo intensity (rho = 0.27, p = .024) were associated with more severe AHI (see Figure 7). Notably, similar correlations with AHI were observed for posterior tongue fat and ultrasound echo intensity. Associations with ultrasound intensity and AHI were slightly reduced in the full sample with or without available MRI in unadjusted (rho = 0.27, p = .015) and adjusted (rho = 0.20, p = .077) analyses, although it is unclear whether the same reductions would have occurred if we had been able to quantify MRI measurements in these additional patients.

Table 2.

Unadjusted and adjusted correlations of ultrasound and MRI with AHI severity among cohort with non-missing MRI measures

Characteristic N Unadjusted Adjusted*
rho P rho P
Ultrasound measures
 Echo intensity 74 0.34 .003 0.27 .024
 Tongue height, mm 70 0.17 .166 0.10 .420
 Tongue width, mm 70 0.31 .009 0.18 .145
MRI measures
 Posterior tongue fat, % 74 0.35 .002 0.28 .020
 Whole tongue fat, % 74 0.42 .0002 0.38 .001
 Tongue height, mm 70 0.21 .087 0.15 .243
 Tongue width, mm 70 0.32 .008 0.19 .119
Open in a new tab

*Partial correlation adjusted for age, BMI, sex and race (white vs. non-white).

Figure 7.

Figure 7.

Open in a new tab

Correlations of ultrasound and MRI-based tongue fat percentages with the apnea-hypopnea index (AHI). Higher values of (A) ultrasound echo intensity and both (B) posterior and (C) whole tongue fat percentages from MRI were associated with more severe AHI, even after controlling for clinical covariates of age, BMI, sex, and race.

Discussion

This study is the first to examine the ability of ultrasound echo intensity to characterize tongue fat. The results of our ex-vivo cow tongue experiment showed a strong positive relationship between increasing fat volume and increased ultrasound echo intensity. Ultrasound echo intensity showed strong correlation to MRI-derived tongue fat percentage among human subjects. In agreement analyses, ultrasound-derived tongue width and height had small to medium standardized limits of agreement compared to MRI measures, suggesting good agreement. Although there were no significant differences in tongue measurements between OSA and non-OSA groups using an AHI cutoff of 5 events/hour, larger values of ultrasound echo intensity and both posterior and whole tongue fat percentage based on MRI were associated with more severe AHI in regression analysis controlling for covariates. Tongue fat alone may contribute less to mild (5–15 events/hour) versus moderate/severe (>15 events/hour) OSA, but future studies in larger samples stratified by OSA severity are needed to better understand this relationship. Overall, results suggest ultrasound is a viable surrogate measure for characterizing the tongue in terms of both fat content and linear dimensions and captures similar associations with OSA severity as tongue fat from MRI when quantified in the same region of the tongue.

Our ultrasound method for calculating echo intensity was limited to a two-dimensional cross-section of the tongue base, while MRI can characterize the entire tongue. The tongue base was the focus of the ultrasound analysis because prior studies have shown it contains the greatest amounts of fat in the tongue [25]. Correlation of echo intensity was strongest with tongue fat percentage of the posterior tongue on MRI, which is the region focused on by the ultrasound. The correlation was less strong when compared to tongue fat of the entire tongue. Given the importance of whole tongue fat percentage for characterizing OSA risk, future studies could develop ultrasound algorithms to capture multiple areas of the tongue to calculate echo intensity that includes more regions of the tongue. Importantly, our results demonstrate excellent correlation between ultrasound and MRI-based tongue fat percentage when applied to similar regions of the tongue, as well as similar associations with severity of AHI.

Echo intensity is an indirect measure of fat content based on interactions of fat and muscle in the tongue. A tissue’s echo intensity is also influenced by other soft tissue components, including connective tissue, nerves, and blood vessels [13]. These can influence the accuracy of echo intensity as a quantitative metric of fat, particularly fibrous and connective tissue which also appear with greater echo intensity on ultrasound [13]. Analysis of cadaveric humans has shown that the majority of the tongue consists of muscle and adipose tissue, with smaller contributions from connective tissue [26]. This tissue makeup may be why the correlation of ultrasound echo intensity of the tongue with MRI tongue fat percentage was stronger than those reported in skeletal muscles of the lower extremities [26]. Ultimately, an inherent limitation of ultrasound is its inability to distinguish between echo intensity from fat or connective tissue, leaving MRI as a superior modality for accurate quantification of tongue fat. However, there are number of benefits of utilizing ultrasound for capturing information on tongue fat.

Strengths of ultrasound as a tongue imaging modality include its ease of use, portability, low cost, and low patient risk. Ultrasound devices could easily be incorporated into the outpatient clinical workflow or during a sleep study, facilitating evaluations of large numbers of patients, and may have future utility as a screening method to stratify patients for OSA risk or expected disease severity, as well as potentially guide treatment options. Ultrasound could also be used as a means of high-throughput phenotyping of the tongue for large-scale genetic studies. Furthermore, future studies in the context of weight loss, bariatric or tongue reduction surgery, or other novel therapies aimed at reducing tongue fat, would be useful to establish the clinical applications for submental ultrasound with regard to serial monitoring of the tongue to evaluate changes in size and echo intensity over time.

Our study had several important strengths. First, we included a relatively large number of participants with and without OSA to demonstrate the strong relationship between ultrasound and MRI measurements. Second, an ex-vivo animal model was used to confirm that increasing fat increases ultrasound echo intensity. Third, ultrasound was performed by a single evaluator using the same ultrasound model and same settings among all subjects to ensure consistency. Fourth, MRI and ultrasound evaluators were blinded to each other’s results to minimize possible bias. Finally, repeated measures on ultrasound demonstrated strong reproducibility of both the measurement and analysis.

There were also some important limitations of our study. While ultrasound analysis being performed by a single evaluator reduces bias in the present study, ultrasound can be operator dependent and future studies will be needed to determine interrater reliability of ultrasound echo intensity and linear measurements. Our study focused on demonstrating a relationship between MRI-based tongue measurements and those obtained using ultrasound. We also did not establish standardized values or normal limits of echo intensity that could be translated for clinical interpretation. An important next step is to evaluate the clinical utility of ultrasound and identify both normal limits and, relatedly, specific cut-points that may be predictive of increased likelihood of OSA. Additionally, future studies should explore the relationship of tongue fat and ultrasound echo intensity to other metrics associated with adiposity or OSA severity (neck circumference, hip/waist circumference, skin fold thickness), which were not obtained in this study. Finally, while relatively large, our sample size was only adequately powered to detect moderate correlations; thus, nonsignificant results for smaller effects should be interpreted cautiously. Future studies in larger number of patients will be important to validate the observed effects and elaborate on differences in tongue characteristics obtained using submental ultrasound, including with respect to the potential utility of ultrasound for distinguishing patients with and without OSA.

Conclusions

Submental ultrasound can characterize the tongue and provides greater ease of use, improved portability, lower cost, and less patient burden compared to MRI. Ultrasound echo intensity can be used with linear measurements to characterize the size and fat content of the tongue, with a more accessible imaging modality than MRI. Moreover, ultrasound echo intensity demonstrated similar associations with severity of the AHI as MRI-based tongue fat percentage in the same region. Thus, ultrasound represents a promising tool for use in the OSA population, where evaluation of tongue fat using a fast and convenient modality can be a valuable tool to better diagnose, phenotype, and manage this disorder.

Funding

Supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under award number P01HL094307.

Disclosure Statement

Financial Disclosure: R.J.S. has received research support from the following: Resmed, Inspire, and CryOSA. He receives royalties for contributions to UptoDate and Merck Manual. All other authors have no financial disclosures to declare.

Nonfinancial Disclosure: R.J.S is on the scientific advisory board for eXciteOSA. All other authors have no non-financial disclosures to declare.

Authors’ Contributions

JLY, AW: study conception and design, data acquisition, interpretation, analysis, and drafting the manuscript. SMS: study conception and design, manuscript revision for important intellectual content. BTK: data interpretation and manuscript revision for important intellectual content. CMS, RJS.: study conception and design, manuscript revision for important intellectual content, and general study supervision.

All authors approved the final manuscript version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.

References

  • 1. Peppard PE, et al. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol. 2013;177(9):1006–1014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Punjabi NM. The epidemiology of adult obstructive sleep apnea. Proc Am Thorac Soc. 2008;5(2):136–143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Schwab RJ, et al. Identification of upper airway anatomic risk factors for obstructive sleep apnea with volumetric magnetic resonance imaging. Am J Respir Crit Care Med. 2003;168(5):522–530. [DOI] [PubMed] [Google Scholar]
  • 4. Dempsey JA, et al. Pathophysiology of sleep apnea. Physiol Rev. 2010;90(1):47–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Kim AM, et al. Tongue fat and its relationship to obstructive sleep apnea. Sleep. 2014;37(10):1639–1648. doi: 10.5665/sleep.4072. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Wang SH, et al. Effect of weight loss on upper airway anatomy and the apnea-hypopnea index. The importance of tongue fat. Am J Respir Crit Care Med. 2020;201(6):718–727. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Lahav Y, et al. Tongue base ultrasound: a diagnostic tool for predicting obstructive sleep apnea. Ann Otol Rhinol Laryngol. 2009;118(3):179–184. [DOI] [PubMed] [Google Scholar]
  • 8. Shu CC, et al. The use of sub-mental ultrasonography for identifying patients with severe obstructive sleep apnea. PLoS One. 2013;8(5):e62848. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Liao LJ, et al. Submental ultrasonography in diagnosing severe obstructive sleep apnea syndrome. J Med Ultrasound. 2016;24(3):107–111. [Google Scholar]
  • 10. Bilici S, et al. Submental ultrasonographic parameters among patients with obstructive sleep apnea. Otolaryngol Head Neck Surg. 2017;156(3):559–566. [DOI] [PubMed] [Google Scholar]
  • 11. Chien CY, et al. Tracking dynamic tongue motion in ultrasound images for obstructive sleep apnea. Ultrasound Med Biol. 2017;43(12):2791–2805. [DOI] [PubMed] [Google Scholar]
  • 12. Manlises CO, et al. Dynamic tongue area measurements in ultrasound images for adults with obstructive sleep apnea. J Sleep Res. 2020;29(4):e13032. [DOI] [PubMed] [Google Scholar]
  • 13. Pillen S, et al. Skeletal muscle ultrasound. Neurol Res. 2011;33(10):1016–1024. [DOI] [PubMed] [Google Scholar]
  • 14. Pillen S, et al. Quantitative skeletal muscle ultrasonography in children with suspected neuromuscular disease. Muscle Nerve. 2003;27(6):699–705. [DOI] [PubMed] [Google Scholar]
  • 15. Watanabe Y, et al. Echo intensity obtained from ultrasonography images reflecting muscle strength in elderly men. Clin Interv Aging. 2013;8:993–998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Zaidman CM, et al. Muscle ultrasound quantifies disease progression over time in infants and young boys with Duchenne muscular dystrophy. Muscle Nerve. 2015;52(3):334–338. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Akazawa N, et al. Muscle mass and intramuscular fat of the quadriceps are related to muscle strength in non-ambulatory chronic stroke survivors: a cross-sectional study. PLoS One. 2018;13(8):e0201789. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Fukumoto Y, et al. Association of physical activity with age-related changes in muscle echo intensity in older adults: a 4-year longitudinal study. J Appl Physiol (1985). 2018;125(5):1468–1474. [DOI] [PubMed] [Google Scholar]
  • 19. Young HJ, et al. Measurement of intramuscular fat by muscle echo intensity. Muscle Nerve. 2015;52(6):963–971. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Landis JR, et al. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159–174. [PubMed] [Google Scholar]
  • 21. Bakdash JZ, et al. Repeated measures correlation. Front Psychol. 2017;8:456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Altman DG, et al. Measurement in medicine: the analysis of method comparison studies. J R Stat Soc Ser Stat. 1983;32(3):307–317. [Google Scholar]
  • 23. Hochberg Y. A sharper Bonferroni procedure for multiple tests of significance. Biometrika. 1988;75(0):800–802. [Google Scholar]
  • 24. Cohen J. Statistical Power Analysis for the Behavioral Sciences. New York: Academic Press; 2013. [Google Scholar]
  • 25. Nashi N, et al. Lingual fat at autopsy. Laryngoscope. 2007;117(8):1467–1473. [DOI] [PubMed] [Google Scholar]
  • 26. Miller JL, et al. Muscle, adipose, and connective tissue variations in intrinsic musculature of the adult human tongue. J Speech Lang Hear Res. 2002;45(1):51–65. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data underlying this article will be shared on reasonable request to the corresponding author.

RESOURCES