点击显示 收起
| ABSTRACT |
|---|
| TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES |
|---|
Key Words: magnetic resonance imaging • obstructive sleep apnea syndrome • upper airway
| INTRODUCTION |
|---|
| TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES |
|---|
Magnetic resonance imaging (MRI) allows visualization and accurate measurement of the upper airway as well as of the various soft tissues and skeleton comprising it (3–6). Using MRI, we have previously shown that children with OSAS and no apparent craniofacial or neurologic disorder have decreased upper airway volume and increased adenoid and tonsillar volume in comparison with control subjects (5).
The three-dimensional relationship between the upper airway, adenoid, and tonsils has not been previously studied in children with OSAS. The aim of the present study was to characterize the differential contribution of the adenoid and tonsils to airway restriction along the upper airway by determining the cross-sectional airway area in planes orthogonal to the upper airway axis. To this end, we used a new methodology, developed on the basis of fuzzy connectedness-based automatic segmentation (7–9), that enabled us to visualize and analyze the upper airway in a correct anatomic orientation as it relates to airflow and further delineate the upper airway changes that occur in children with OSAS.
| METHODS |
|---|
| TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES |
|---|
Control Subjects
Twenty children with normal growth and development were matched to subjects with OSAS by age, sex, ethnicity, weight, and height. Control subjects were selected from among patients who underwent head MRI at the Children's Hospital of Philadelphia for other medical indications. Exclusion criteria included the following: (1) likelihood of OSAS (scores of -1 or more; assessed by a standard questionnaire [10]), (2) evidence of a brain tumor or a seizure disorder requiring therapy, (3) genetic disorders associated with any craniofacial anomaly, and (4) chronic respiratory disease such as asthma or bronchopulmonary dysplasia.
Overnight Polysomnography
For subjects with OSAS, polysomnography was performed 0–4 weeks before MRI. Subjects were studied in the Sleep Disorders Center at the Children's Hospital of Philadelphia. Scoring of respiratory variables was performed on the basis of standards set by the American Thoracic Society and previously published data on children (11, 12). Flow was measured with an oral/nasal thermistor and a nasal end-tidal PCO2 catheter. We used the definition of obstructive apnea as absence of oral/nasal thermistor signal for at least two respiratory cycles associated with out-of-phase movement of the rib cage and abdomen. Hypopnea was defined as a decrease of 50% or more in oral/nasal thermistor signal and a concurrent fall of 4% or more in basal oxygen saturation. Sleep stages were determined by the criteria of Rechtschaffen and Kales (13).
Sleep Questionnaire
A questionnaire regarding symptoms of sleep-disordered breathing based on the questionnaire developed by Brouillette and coworkers (10) was used to assess the likelihood of OSAS in control subjects and subjects with OSAS. On the basis of the questionnaire, no subject with a score less than -1 would be expected to have OSAS; a score between -1 and 3.5 is considered indeterminate, and a score greater than 3.5 is considered highly predictive of obstructive sleep apnea.
Magnetic Resonance Imaging
MRI studies were performed in the Department of Radiology at the Children's Hospital of Philadelphia. All studies were performed while subjects were sedated. Sedation consisted of intravenous pentobarbital (2-mg/kg increments) until sleep was achieved. A maximum of three doses or 200 mg was administered. All subjects were monitored continuously by pulse oximetry and were observed by an intensive care unit attending physician until recovery (about 1 hour).
All MR images were obtained as part of a larger ongoing study (unpublished data), using a comprehensive MR protocol. Images of 15 of the 40 subjects in this study (9 subjects with OSAS and 6 control subjects) were used in a previous study (5), and were now processed for the first time according to the new methodology (see below) to further delineate airway anatomic characteristics.
MRI was performed with a 1.5-T Vision System (Siemens, Iselin, NJ). Images were acquired with an anterior–posterior volume head coil. The patient's head was positioned supine in the soft tissue Frankfort plane (tragus of the ear to orbital fissure) perpendicular to the table. Sequential T2-weighted spin-echo axial sections were obtained, spanning from the orbital cavity to the larynx. Mean acquisition time was 2 minutes, spin-echo repetition time (TR) = 650 milliseconds, echo time (TE) = 14 milliseconds, 192 x 256 matrix, slice thickness 3 mm with distance factor 0, one acquisition, field of view (FOV) = 20 to 24 cm, rectangular FOV 6/8.
Image Processing and Anatomic Measurements
The MR images were transferred to a SUN workstation. Images were processed and segmented automatically to compute the airway including its surface description, centerline, and volume , using a software program we have developed utilizing fuzzy connectedness segmentation (7–9) and based on 3DVIEWNIX (14).
|
Upper airway cross-sectional area measurements.
Cross-sectional areas at planes orthogonal to the centerline were computed every 0.2 mm after interpolation, filtering, and thresholding of the original axial slices (9). We measured the mean and minimal cross-sectional area of the total airway; the mean cross-sectional area in regions adjacent to the adenoid, adjacent to the overlap region between the adenoid and tonsils, and adjacent to the tonsils; and the mean and minimal cross-sectional area for 10 consecutive segments at 10% increments of the centerline.
Upper airway volume measurements.
The total airway volume was computed as the product of centerline length and mean cross-sectional area. Similarly, regional airway volumes adjacent to the adenoid, adjacent to the overlap region between the adenoid and tonsils, and adjacent to the tonsils were computed as the product of mean airway cross-sectional areas along these regions and the corresponding centerline lengths. Finally, volumes for 10 consecutive segments at 10% increments of the centerline were computed.
Data Analysis
Mean and standard deviation were used to summarize continuous variables. For comparisons between the groups for MRI data, demographics, and questionnaire data, we used a two-tailed unpaired t test, the Wilcoxon rank test, or the 2 test as appropriate. A p value of 0.05 or less was considered to indicate statistical significance.
| RESULTS |
|---|
| TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES |
|---|
|
Sleep Questionnaire
All control subjects had an apnea score of less than –1, indicating unlikelihood of obstructive sleep apnea (10), and as a group had a mean apnea score of -3.3 ± 0.6. This score was significantly lower than the apnea score noted in the OSAS group of 2.9 ± 1.5 (p < 0.0005).
Magnetic Resonance Imaging
Representative three-dimensional display of the upper airway, its centerline, and cross-sectional images orthogonal to the centerline at levels of the adenoid, the region of overlap between adenoid and tonsils, tonsils, and the superior portion of the epiglottis of a control subject and a subject with OSAS are presented in and , respectively.
|
Regional analysis.
Analysis of the airway along the adenoid, the region of overlap between adenoid and tonsils, and tonsils in subjects with OSAS and control subjects is shown in . Accordingly, we noted a significantly smaller mean cross-sectional area in subjects with OSAS in all three regions and significantly smaller mean volume and a longer mean airway centerline in regions adjacent to the adenoid and where adenoid and tonsils overlap, but not in the region adjacent to the tonsils. The smallest mean cross-sectional area and volume were noted for both OSAS and control subject groups in the airway region where adenoid and tonsils overlap.
|
|
| DISCUSSION |
|---|
| TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES |
|---|
Some methodological issues need initial comment. The upper airway has a complex shape formed by various tissues with heterogeneous compositions that surround the air column. This, along with airway boundary motion, results in some MR signal gradation and blurring. This was true in the present study when data were collected during a 2-minute period throughout numerous respiratory cycles. Fuzzy connectedness is a computational method of object segmentation that takes into account the effects of tissue heterogeneity, imaging noise, and blurring. This method determines voxels that comprise an object by assigning weights to each voxel on the basis of its spatial nearness, similarity of intensity, and pathwise relationship to all other voxels. The resultant fuzzy object is a pool of voxels with a membership value that represents its strength as a member of the object (7, 8).
Fuzzy connectedness is a robust tool for airway segmentation. It achieves an intraoperator and interoperator variation of less than 0.87% and the accuracy of volume and delineation was found to be 97% as compared with expert manual segmentation (9). In addition, when applied to the airway, fuzzy connectedness was performed in less than 10% of the time required for manual segmentation (9).
For the most reliable comparison of the OSAS group with the control group, we performed a case–control study and matched each subject with OSAS by age, height, weight, sex, and ethnicity. All could influence the airway by affecting size, shape, and function of the surrounding tissues, including muscle, lymphoid, fat, and bone (5, 15, 16). Our control subjects did not undergo polysomnography and were screened only by history and a standardized questionnaire. No obstructive events were observed in the control subjects during imaging.
To obtain optimal MR images in young children, mild sedation to avoid body movement is necessary and is routinely used in our institution. Sedation can reduce upper airway muscle activity compared with wakefulness and could have affected our airway measurements (17). It is possible that sedation had a bigger effect on upper airway muscle activation in subjects with OSAS compared with control subjects, amplifying the differences between the groups in our study. Moreover, sedation could have caused partial or complete airway obstruction during data acquisition, leading to increased motion and resulting in more blurring of our images in the OSAS group. Hence, the limitation of sedation in our study should be recognized.
In the present study we confirm our previous finding (5), suggesting that the size of the upper airway in children with OSAS is smaller compared with normal children. However, the current study was more comprehensive because it was designed to identify the actual site(s) of restriction along the upper airway path and the role of the adenoid and tonsils in airway restriction in these subjects, using a new methodology.
In the previous study, we used T1 images for airway analysis. The airway measurements included a single axial cross-sectional area at the midtonsillar level, an airway volume derived from sequential 3-mm axial slices, and a single nasopharyngeal cross-sectional area obtained from a midsagittal image. All measurements were made manually with the program VIDA and no airway reconstruction was performed.
In the present study, we reconstructed the airway using T2 images with a new algorithm in a program 3DVIEWNIX. This method is a multistep process, including nonmanual fuzzy connectedness delineation, interpolation, filtering, and thresholding that results in a three-dimensional display of the airway, centerline, area profile, and cross-sectional images (9). This method enables accurate representation of the airway as it relates to airflow and surrounding tissues and is significantly more efficient than manual segmentation.
In the present study, we found that the upper airway cross-sectional area varies along the airway centerline in a similar form in both normal children and in children with OSAS, with the OSAS group being about one-half smaller in the upper two-thirds of the airway. In both groups, about 20 to 60% of the airway represents a region where adenoid and tonsils overlap ( bottom); this region corresponds to the lowest mean and minimal airway cross-sectional areas. It is possible that airway obstruction in children with OSAS occurs in this region. However, our study could not prove this speculation because of the relatively long MR acquisition time, and further investigation in another study using faster MR sequences will be needed to visualize dynamic changes of the airway during respiration.
Isono and coworkers (18) assessed the collapsibility and minimal cross-sectional area of the passive airway in children with OSAS (age, 7.6 ± 3.5 years). These children were studied under general anesthesia and measurements were performed by endoscopy at discrete levels including the adenoid, soft palate, tonsils, and tongue. These investigators noted the mean highest closing pressure and minimal cross-sectional area to occur mostly at the levels of adenoid and soft palate. These findings suggesting that higher airway segments are more involved in children with OSAS support our findings showing restriction in the upper two-thirds of the airway.
Although the total centerline length was similar in OSAS and control groups, in regions adjacent to the adenoid and where the adenoid and tonsils overlap, the lengths were significantly longer in the OSAS group (22.6 ± 5.5 versus 19.5 ± 3.8 mm, p < 0.05, and 15.4 ± 4.5 versus 12.5 ± 3.1 mm, p < 0.05, respectively). This could be related to the increased extension of lymphoid tissue in OSAS and the more tortuous path along the centerline in these narrowed regions. In addition, segmental analysis along the entire centerline suggests that the upper airway of children with OSAS is restricted compared with control subjects over the initial 60–70% of its length. The narrowing is not a discrete region adjacent to either the adenoid or tonsils, but rather occurs in a continuous fashion along both. These observations suggest that flow resistance is higher in these regions, and complete obstruction during inspiration in sleep may be more likely where narrowing occurs and high negative pressures are developed.
Interestingly, segmental analysis suggests that airway measurements below the region of maximal restriction (i.e., region of overlap between adenoid and tonsils) are similar in both groups . This could explain to some extent why clinical assessment of the tonsils and airway that are below this region of restriction do not predict the existence or severity of OSAS (19–21).
In summary, we used MRI to analyze the size of the upper airway in children with OSAS. Our results suggest that the upper airway in children with OSAS is significantly smaller with respect to airway volume, mean airway cross-sectional area, and minimal cross-sectional area compared with matched control subjects. We also noted that the upper airway in children with OSAS is restricted along the initial 60–70% of its length and most affected in regions where adenoid and tonsils overlap.
| REFERENCES |
|---|
| TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES |
|---|