Published ahead of print on October 20, 2005, doi:10.1164/rccm.200507-1148OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200507-1148OC
Quantitative Upper Airway Imaging with Anatomic Optical Coherence TomographyOptical+Biomedical Engineering Laboratory, School of Electrical, Electronic, and Computer Engineering, and School of Anatomy and Human Biology, University of Western Australia, Crawley; and West Australian Sleep Disorders Research Institute, Department of Pulmonary Physiology, Sir Charles Gairdner Hospital, Nedlands, Western Australia, Australia Correspondence and requests for reprints should be addressed to Julian Armstrong, B.Sc., B.E. (Hons), Optical+Biomedical Engineering Laboratory, School of Electrical, Electronic, and Computer Engineering, University of Western Australia, 35 Stirling Highway, Crawley, Western Australia 6009. E-mail: julian-a{at}ee.uwa.edu.au
Background: Measurements of upper airway size and shape are important in investigating the pathophysiology of obstructive sleep apnea (OSA) and in devising, applying, and determining the effectiveness of treatment modalities. We describe an endoscopic optical technique (anatomic optical coherence tomography, aOCT) that provides quantitative real-time imaging of the internal anatomy of the human upper airway. Methods: Validation studies were performed by comparing aOCT- and computed tomography (CT)–derived measurements of cross-sectional area (CSA) in (1) conduits in a wax phantom and (2) the velo-, oro-, and hypopharynx during wakefulness in five volunteers. aOCT scanning was performed during sleep in one subject with OSA. Results: aOCT generated images of pharyngeal shape and measurements of CSA and internal dimensions that were comparable to radiographic CT images. The mean difference between aOCT- and CT-derived measurements of CSA in (1) the wax phantom was 2.1 mm2 with limits of agreement (2 SD) from –13.2 to 17.4 mm2 and intraclass correlation coefficient of 0.99 (p < 0.001) and (2) the pharyngeal airway was 14.1 mm2 with limits of agreement from –43.7 to 57.8 mm2 and intraclass correlation coefficient of 0.89 (p < 0.001). aOCT generated quantitative images of changes in upper airway size and shape before, during, and after an apneic event in an individual with OSA. Conclusions: aOCT generates quantitative, real-time measurements of upper airway size and shape with minimal invasiveness, allowing study over lengthy periods during both sleep and wakefulness. These features should make it useful for study of upper airway behavior to investigate OSA pathophysiology and aid clinical management.
Key Words: optical coherence tomography • sleep apnea • upper airway anatomy Measurements of upper airway size and shape are important in investigating the pathophysiology of obstructive sleep apnea (OSA) and in devising, applying, and determining the effectiveness of treatment modalities (1). However, a persisting limitation is the absence of an imaging technique that satisfactorily provides safe, accurate, and repetitive measurements of upper airway dimensions during wakefulness and sleep. Radiographic computed tomography (CT) and fluoroscopy involve potentially hazardous ionizing radiation. Magnetic resonance imaging (MRI) is expensive, noisy, claustrophobic, and incompatible with metallic probes and catheters. Neither modality is suitable for lengthy evaluations or evaluations during sleep (2). Ultrasound is not feasible because of poor transducer-air coupling, an unavoidable problem in air-filled organs. Fiber-optic nasoendoscopy has been used to visually examine the anatomic features and structure of the airway, but its quantification of airway dimensions is indirect and subjective (3). Measurement of airway caliber using acoustic reflection is not possible distal to the site of collapse, because the acoustic waves will not penetrate a collapsed airway (4). In this article, we describe an endoscopic optical technique that generates quantitative, real-time images of the upper airway that enable accurate determination of shape and size. Briefly, an optical probe is placed inside a catheter, which is inserted via the nares to the level of the midesophagus. Rotation of the probe within the catheter provides a 360° profile of surrounding tissue. The optical probe can be systematically moved within the catheter, allowing the upper airway to be scanned at multiple sites without stimulating the airway mucosa. The technique is adapted from optical coherence tomography (OCT) (5), a medical imaging modality. OCT is an optical sectioning microscopy modality that has been used to image subsurface tissue morphology in fields including ophthalmology (6), dermatology (7), vascular medicine (8), gastroenterology (9), and urology (10). Our adaptation involves a substantial increase in its axial distance range to allow macroscopic imaging of upper airway internal anatomy. We refer to this adaptation as "anatomic OCT" (aOCT) to distinguish it from the existing microscopic, subsurface OCT techniques (11). This article presents the operational capabilities of the system, including data regarding its validity and applicability to quantitative measurement of human upper airway size and shape.
Baseline in vitro validation studies were performed by comparing aOCT-derived measurements of cross-sectional area (CSA) with those derived from (1) circular holes of known size in a wooden template and (2) CT-derived measurements of holes with varying shape and size cut in a wax phantom. After these measurements, in vivo studies were performed to (1) compare aOCT-derived measurements of upper airway shape and size at the levels of the velo-, oro-, and hypopharynx with those obtained from simultaneous CT scans and (2) determine the reproducibility, and intra- and interobserver variability of aOCT-derived measurements of upper airway shape and size. Additional in vivo measurements were performed to demonstrate the feasibility of undertaking such measurements to determine pharyngeal compliance and, in one patient with OSA, during sleep. Approvals for all measurements in human volunteers were obtained from the Human Research Ethics Committees of Sir Charles Gairdner Hospital and the University of Western Australia. Informed consent was obtained from all subjects.
Description of aOCT The system operates by directing a light beam perpendicular to the catheter. The distance between the probe head and the air–tissue interface of the airway wall is determined from the reflected light using a low-coherence optical interferometer. The probe rotates at 1.25 Hz to capture quantitative cross-sectional images of the upper airway lumen, in much the same way that a radar system captures an image. A customized computer program controls a motorized translation stage that allows the probe head to be precisely rotated and precisely translated to various levels within the pharynx to record cross-sections of interest. These tomograms can be viewed graphically in real time on a personal computer or reconstructed in the form of a video. Additional details regarding the technique can be found in the online supplement.
Validation Studies
In vivo validation studies.
aOCT versus CT.
Reproducibility and intra- and interobserver variability.
To assess intra- and interobserver variability of aOCT-derived measurements of pharyngeal dimensions, 10 images of each of the velo-, oro-, and hypopharynx were randomly selected from a database containing images from 10 subjects and analyzed by two observers on two separate occasions (separated by
Feasibility Studies
aOCT scanning during sleep.
Analyses Regression analyses were performed with the least-squares method. Measurements of pharyngeal compliance were compared between groups at each site using two-way repeated measures analysis of variance. A Holm-Sidak test was applied for all post-hoc comparisons. The Bland-Altman test (14) and the intraclass correlation coefficient (15) were used to evaluate the agreement between aOCT and CT-derived measurements of CSA from the wax phantom and pharyngeal dimensions, and to evaluate reproducibility and intra- and interobserver variability of aOCT-derived measurements.
Validation Studies In vitro bench studies. The percentage errors in aOCT-derived measurements of CSA for five holes drilled in a template constructed from wood with radii of 5, 10, 15, 20, and 25 mm were 0.3, 0.6, 0.3, –2.7, and –1.1%, respectively. The mean error was –0.5%, or 0.1 cm2, which corresponded to a mean error in radius of 0.07 mm. The differences in CSA of 23 conduits of varying shape and size cut in a wax phantom are demonstrated in a Bland-Altman plot (Figure 1). The average CSA obtained using aOCT and CT was 610 ± 500 and 608 ± 499 mm2, respectively. The mean difference between the measurements of CSA obtained with the two techniques was 2.1 mm2 with limits of agreement (2 SD) from –13.2 to 17.4 mm2 and an intraclass correlation coefficient of 0.99 (p < 0.001). The mean differences between measurements of anteroposterior and lateral diameters were small, being 0.31 and 0.54 mm, respectively. The intraclass correlation coefficient was 0.99 for measurements of anteroposterior (p < 0.001) and lateral diameters (p < 0.001).
In Vivo Validation Studies Pullback scans. Figure 2 shows lateral and anteroposterior longitudinal reconstructions from a pullback scan with six cross-sectional images at locations spaced at intervals along the pharynx. A movie created from the pullback scan (see Video E1) is available in the online supplement. A number of anatomic features are indicated on the images. The dark lines in the images are the air-tissue boundaries of the lumen; for the cross-sections these are traced out by the aOCT system during a single rotation. The tissue–air boundaries in the figure appear fuzzy, because aOCT also detects subsurface reflections, but the air–airway wall interface is sharp, allowing accurate surface location.
aOCT versus CT. Figure 3 shows representative examples of simultaneous aOCT and CT scans at each of the three levels in one healthy subject without OSA. A Bland-Altman plot of the pooled difference in CSA using the two techniques in five healthy subjects is shown in Figure 4. The average CSA obtained using aOCT and CT was 133.2 ± 61.4 and 147.4 ± 59.7 mm2, respectively. The mean difference between measurements obtained with the two techniques was 14.1 mm2 with limits of agreement (2SD) from –43.7 to 57.8 mm2. The mean difference in measures of anteroposterior and lateral diameters was 0.7 and 0.4 mm, respectively, with limits of agreement from –3.5 to 4.8 mm and –6.8 to 7.6 mm, respectively. The intraclass correlation coefficients for measurements of CSA and anteroposterior and lateral diameters were 0.89, 0.66, and 0.82, respectively (p < 0.05 for all). Data for measurements obtained at each of the three sites are shown in Table 1 and Tables E2 and E3.
Reproducibility. To assess reproducibility, five individuals were scanned with aOCT on two separate occasions, and measurements of pharyngeal dimensions obtained at the same locations within the velo-, oro-, and hypopharynx using the same observer to derive the measurements. The average CSA on the first testing occasion was 161.2 ± 43.6, 226.9 ± 42.8, and 210.3 ± 91.4 mm2, respectively, and 167.9 ± 60.8, 242.6 ± 45.7, and 188.1 ± 93.3 mm2 on the second occasion. The mean difference between the two testing days was 6.7, 15.7, and 22.2 mm2, respectively, and the intraclass correlation coefficients for measurements obtained on the two testing days were 0.75, 0.70, and 0.97, respectively (p < 0.01 for all; Table 1). The mean difference and intraclass correlation coefficients between duplicate measurements of anteroposterior and lateral diameters are reported in Tables E2 and E3.
Intraobserver variability.
Interobserver variability.
Feasibility Studies
aOCT scanning during sleep. Figure 6 demonstrates changes in upper airway size and shape before, during, and after airway collapse (i.e., an apneic event) in an individual with OSA. Images were collected in the velopharynx while the subject was in stable stage 2 sleep. The 10 selected aOCT cross-sectional images in Figures 6a–6j show the airway changing size during normal unobstructed breathing, being smaller during inspiration than expiration (a–d); the airway narrowing (e) then collapsing during a 5-s apnea (f); collapsing during a 12-s apnea (g, h); and restoration of an open airway after momentary arousal from sleep (i, j). Video E3 shows aOCT data of the apneic event.
This study reports on a new optical technique that can provide quantitative real-time imaging of the internal anatomy of the human upper airway. The impetus for this work has been the long-recognized need to quantitatively evaluate changes in human upper airway size and shape during sleep. Such measurements are of particular value in understanding the pathophysiology of OSA and in devising, applying, and determining the effectiveness of treatment modalities (1). In recent years, upper airway size and shape have been measured by a number of techniques, including CT (16–18), fluoroscopy (19, 20), MRI (21–24), nasoendoscopy (25–28), and acoustic reflection (4, 29, 30). Although these studies have demonstrated important differences in pharyngeal shape and size between individuals with and without OSA, in terms of baseline anatomy (4, 16, 18, 22, 23, 29, 30) and dynamic behavior (17, 19–21, 24–28), the routine use of these technologies for study of the upper airway on repeated occasions and over prolonged periods in individuals is limited. We present aOCT as a useful addition to these existing technologies. Its minimal invasiveness, quietness, comfort, and lack of potentially hazardous effects (such as ionizing radiation) make it suitable for study of pharyngeal size and shape over prolonged periods during both sleep and wakefulness. The in vitro measurements illustrate the precision of the method in obtaining measurements of shape and size. The differences between aOCT and CT-derived measures of anteroposterior and lateral dimensions were very small (0.4 mm on average), and the high intraclass correlation coefficients demonstrate that the two techniques provide comparable measurements. CSA measurements tended to be slightly larger when measured using aOCT (by 2.1 mm2 on average), although the high intraclass correlation coefficients demonstrate that the two techniques show good agreement. In each of the in vivo studies presented in this article, aOCT scanning was commenced with a pullback scan (Figure 2). This was achieved by systematically retracting the optical probe from the upper esophagus to the nasal cavity while the subject was breathing quietly. There is no sensation of movement of the probe because it is retracted within a fixed catheter, thereby avoiding stimulation of the airway mucosa. Also, because the mass of the rotating fiber probe is exceedingly low and substantially less than the plastic catheter, significant mechanical vibration does not transfer from the probe to the catheter. The benefit of this was demonstrated in the one subject who was studied during sleep (see Figure 6), where the probe was moved to various sites in the upper airway without causing arousal. On completion of the pullback scan, each resulting data set was then used to construct a movie (see Video E1). Although the movie constitutes a useful dataset in its own right, it can also be used to identify specific sites for further investigation. In such circumstances, the optical probe is moved to those sites and continuous scanning undertaken to monitor the upper airway during steady-state conditions, state changes (e.g., changes in sleep stage or posture), or imposed changes (e.g., changes in airway pressure). aOCT generated images of pharyngeal cross-sections that were comparable to images obtained by radiographic CT. The differences between aOCT- and CT-derived measures of anteroposterior and lateral dimensions were small (0.8 mm on average) at all levels. CSA measurements tended to be smaller with aOCT, particularly at the level of the hypopharynx. The most likely contributors to this difference were the misalignment between the aOCT and CT scan planes (caused by restricted movement of the CT gantry or occasional head/neck movement between the time of the CT scout scan and the cross-sectional scan), and obscuration of small parts of the airway wall during aOCT scanning (particularly at the hypopharynx; Figure 3). This obscuration, which is a limitation of the aOCT technique (see METHODOLOGIC CONSIDERATIONS), may also have contributed to the larger difference in CSA of the hypopharynx relative to other levels when individuals were measured on two separate occasions. Other contributors to this variability include position of the tongue and soft palate and differences in head and neck position, although the latter were minimized as far as possible between the two testing occasions.
Methodologic Considerations aOCT has the capacity to continuously measure changes in airway dimensions under a variety of conditions. Such changes can be seen during tidal breathing in each of the supplementary videos, and before, during, and after an apneic event in an individual with OSA (Figure 6). The images obtained from this individual show the potential of the technique to describe the patterns of change in airway shape and size that lead up to upper airway collapse. In this individual, the upper airway narrowed and then closed during expiration (e.g., frames e and f), consistent with previous analyses of airflow data (36), nasoendoscopy images (37), and CT images (38). When CSA is related to changes in airway pressure, a measure of pharyngeal compliance can be derived as shown in Figure 5. The finding that, in individuals with OSA, the pharynx was smallest and least compliant at the level of the velopharynx is consistent with a previous CT-based study (39). Although the 1.25-Hz rotation rate of the aOCT system is comparable with the acquisition rates of most CT and MRI scanners, it is not sufficient to track changes in airway caliber when breathing rates are rapid. This can be allowed for by synchronizing the aOCT images with other respiratory data, such as flow or volume change, and multiplexing images obtained over several breaths, so that the composite image is obtained relative to a particular phase of respiration. The need to do this would be obviated by increasing the rotation rate by at least a factor of three, which is an intended technological development. A limitation of the technique is its capacity to view the complete circumference of the airway at all sites in all individuals. In some individuals, on some occasions, the probe may be positioned adjacent to an airway wall that is sufficiently concave or irregular to obscure parts of the airway wall from the probe's light beam. An example of this is the hypopharyngeal region (see Figures 2 and 3), in which the epiglottis can shadow the lower part of the retroglossal pharyngeal wall. In our experience, partial airway profiles occur most commonly at this level, but rarely in the velopharynx, where such gross irregularity of the airway is uncommon (see Figures 2 and 3). Where obscuration does limit measurement of parameters such as total pharyngeal area at that specific site, other useful parameters such as lateral extent of the airway are usually available. Depending on the particular airway and location of the catheter, the scan range required to fully capture the airway cross-section could exceed a 26-mm radius (i.e., 52 mm total cross-section width), the current limit for our system. This is not a major limitation because individuals with OSA typically have either smaller pharynxes overall or a segment of the pharynx, which is narrower than average (16, 23). Nevertheless, it is possible to modify the instrument to extend the range in the future. A challenge to imaging dimensions of the airway using aOCT is that there is longitudinal motion of the airway walls. It is possible, therefore, that the actual level of the structures that a fixed probe is measuring may vary over time as the airway structure moves in a ventrocaudal direction. Such movement was recently demonstrated by Liao and colleagues (40), who used cephalometry to document pharyngeal motion during a Mueller maneuver in a group of patients with sleep-disordered breathing. They showed that, relative to measurements made at end-expiration, a Mueller maneuver increased the length of the pharyngeal airway by 2% in patients with mild OSA and by 6% in patients with moderate to severe OSA. The longitudinal and transmural forces generated during a wakeful Mueller maneuver most likely represent an extreme load on the upper airway, out of proportion to those produced by tidal breathing or by an obstructive event during sleep. Nevertheless, the effect of such longitudinal motion of the airway walls on aOCT-derived measurements of pharyngeal shape and size will need to be considered, perhaps by relating measurements to anatomical landmarks (such as the epiglottis tip).
Future Applications
The authors thank the following individuals for their support and contributions to this work: Kathleen Maddison and Kelly Shepherd from the Department of Pulmonary Physiology, Sir Charles Gairdner Hospital; Peter Muir, Neil Hicks, and Associate Professor Vincent Low from the Department of Radiology, Sir Charles Gairdner Hospital; Alexandre Paduch, Stefan Schwer, and Jonathan Ng from the School of Electrical, Electronic and Computer Engineering, University of Western Australia.
Supported by the Medical Research Foundation of Western Australia and the Australian Health Management Group and by the National Health and Medical Research Council (Australia; development grant no. 303319). These funding contributors had no role in study design, data collection, data analysis, data interpretation, or writing of the report. PRE is supported by a National Health and Medical Research Council (Australia) R. Douglas Wright Fellowship (no. 294404). This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Originally Published in Press as DOI: 10.1164/rccm.200507-1148OC on October 20, 2005 Conflict of Interest Statement: None of the authors have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form July 26, 2005; accepted in final form October 14, 2005
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
24 h). On each occasion, dimensions were calculated from aOCT images obtained at the same level at each of three identical locations (velo-, oro-, and hypopharynx) within the upper airway. 
