Magnetic resonance imaging of obstructive sleep apnea in children
Sleep-disordered breathing has a spectrum of severity that spans from snoring and partial airway collapse with increased upper airway resistance, to complete upper airway obstruction with obstructive sleep apnea during sleeping. While snoring occurs in up to 20% of children, obstructive sleep apnea affects approximately 1–5% of children. The obstruction that occurs in obstructive sleep apnea is the result of the airway collapsing during sleep, which causes arousal and impairs restful sleep. Adenotonsillectomy is the first-line treatment of obstructive sleep apnea and is usually effective in otherwise healthy nonsyndromic children. However, there are subgroups in which this surgery is less effective. These subgroups include children with obesity, severe obstructive sleep apnea preoperatively, Down syndrome, craniofacial anomalies and polycystic ovarian disease. Continuous positive airway pressure (CPAP) is the first-line therapy for persistent obstructive sleep apnea despite previous adenotonsillectomy, but it is often poorly tolerated by children. When CPAP is not tolerated or preferred by the family, surgical options beyond adenotonsillectomy are discussed with the parent and child. Dynamic MRI of the airway provides a means to identify and localize the site or sites of obstruction for these children. In this review the authors address clinical indications for imaging, ideal team members to involve in an effective multidisciplinary program, basic anesthesia requirements, MRI protocol techniques and interpretation of the findings on MRI that help guide surgery.
KeywordsAdenoids Children Laryngomalacia Lingual tonsils Macroglossia Magnetic resonance imaging Obstructive sleep apnea Surgery
Magnetic resonance imaging has been increasingly used to help localize and diagnose the site or sites of obstruction in children with obstructive sleep apnea that persist despite previous adenotonsillectomy. Donnelly  in 2005 described the protocol for using dynamic MRI for this purpose. The current review article serves to update this protocol.
Sleep-disordered breathing has a spectrum of severity that spans from snoring and partial airway collapse with increased upper airway resistance, to complete upper airway obstruction with apnea during sleeping. Snoring can be present with all levels of severity of obstructive sleep apnea and is the symptom that is most indicative of possible obstructive sleep apnea . At its most severe, sleep-disordered breathing causes multiple obstructive episodes each hour resulting in arousal that disturbs restful sleep. When obstructions occur with associated drops in the oxygen saturation or an arousal from restful sleep, then obstructive sleep apnea is present.
Obstructive sleep apnea affects 1–5% of the pediatric population, although it can occur with much higher frequency in children with syndromes and craniofacial abnormalities . Untreated pediatric obstructive sleep apnea can result in cognitive and behavioral problems that can affect academic performance as well as learning [2, 3, 4, 5]. Untreated obstructive sleep apnea can also have more serious complications of pulmonary hypertension, systemic hypertension and cor pulmonale.
The polysomnogram, or sleep study, is the standard by which obstructive sleep apnea and other sleep disorders are diagnosed and severity determined . Pediatric polysomnogram scoring commonly uses a 3% desaturation criterion, unlike adult scoring, which often uses a 4% desaturation criterion [3, 4]. In addition, pediatric obstructive sleep apnea severity definitions differ from those in adults. Severity is determined using the apnea hypopnea index, which is the number of obstructive apneas and hypopnea episodes that occur during sleep, divided by the total sleep time. In children, normal is defined as <1 event per hour. Mild obstructive sleep apnea is from 1 event to <5 events per hour. Moderate obstructive sleep apnea is 5 events to <10 events per hour, and severe obstructive sleep apnea occurs if there are 10 or more events per hour .
In general, first-line treatment for obstructive sleep apnea in children is adenotonsillectomy, and this is effective in most healthy nonsyndromic children. However in a large multicenter study of 578 “normal” children undergoing adenotonsillectomy, the average apnea hypopnea index decreased from 18.2 events/h to 4.1 events/h, and only 27% were “cured” (apnea hypopnea index < 1/h) after adenotonsillectomy . Significant risk factors for persistent obstructive sleep apnea after adenotonsillectomy include obesity, severe obstructive sleep apnea preoperatively, and asthma in nonobese children .
In children who don’t fully respond to adenotonsillectomy, the next line of treatment is frequently long-term continuous positive airway pressure (CPAP), but this is poorly tolerated by children because of its discomfort. Additionally, the contribution of CPAP to disruption of parental rest can contribute to poor adherence . These children with persistent obstructive sleep apnea, poor adherence to CPAP, and a willingness to entertain surgical options are best served by dynamic cine MRI of the upper airway.
The majority of pediatric institutions do not routinely use MRI for treatment planning for persistent obstructive sleep apnea. At institutions that do have such a program, orders for these studies are generally driven by an otolaryngologist who is interested in offering surgical options to families of affected children and who is familiar with the use of MRI in guiding treatment decision-making . A successful program also requires implementation of a standard MRI sleep study protocol (Online supplementary material). This requires an anesthesiologist who is willing to take the lead in providing general anesthesia in this group of children with critical airways . The anesthesiologist must be comfortable in providing anesthesia to a child with a compromised airway without an artificial airway in place.
Magnetic resonance imaging technique
Cine imaging is important for evaluating obstructive sleep apnea. This dynamic exam can show both the degree and the direction of airway collapse, e.g., anterior–posterior (AP) or circumferential. It also shows ways in which children are sometimes compensating or reacting to their periods of obstruction with tongue and jaw thrusting. Dynamic airway imaging is performed using a single-slice multi-phase gradient echo sequence acquired in the axial and sagittal planes. Imaging parameters are selected to obtain as high a spatial resolution as possible (approximately 1.68 mm × 2.62 mm × 5.00 mm in frequency, phase and slice, respectively) while maintaining a target temporal resolution of 300–400 ms/image over 30–35 s (to capture roughly 90 dynamic phases). The repetition time (TR) and echo time (TE; ideally fat and water in phase) are set accordingly, with the flip angle then selected to maximize signal-to-noise ratio. Representative acquisition parameters for an axial dynamic airway acquisition at 1.5 T are FOV=215 × 210 mm (frequency x phase), matrix (frequency x phase)=128 × 82, TR/TE=5.49/3.24 ms, flip angle=15°, acceleration factor=1.5; receiver bandwidth=434 Hz/pixel, signal averages=1.
Cine imaging is performed for approximately 30 s at each level with the rate of imaging of about three images per second, but faster is better. The rate of imaging and quality depends on the field strength, gradient strength, slew rate, matrix and signal-to-noise acceptability with acceleration techniques such as parallel imaging. Newer k-space sampling trajectories, such as compressed sensing, are coming available and are expected to accelerate the temporal resolution.
Axial and sagittal short tau inversion recovery (STIR), or T2-weighted MRI with fat saturation, are very good for visualizing palatine, adenoid and lingual tonsillar tissues, as well as lymphoid tissues and edema (Fig. 1), because these structures appear bright on a dark background. The use of a radial acquisition such as propeller or multivane is robust against motion and blurring. Typically these images are obtained with a FOV=20–24 cm, matrix=240 × 128, and slice thickness=5 mm. T1-weighted imaging is optional but can be performed as a routine if desired.
Some contraindications to MRI do exist. Screening for electronic devices such as pacemakers, defibrillators and vagal nerve stimulators should be routine for MRI because they are contraindications. Although the newest generation of hypoglossal nerve stimulators might be MRI-compatible, the older devices are not, and this must be clarified prior to starting the procedure.
If necessary, all sequences can be performed with the supplemental airway support (CPAP or nasal trumpet), including the sagittal and axial cine imaging. Then the supplemental airway support is removed and the sagittal and axial cine imaging is repeated after removal of the airway support so that the motion of the airway can be evaluated in its natural state. Airway support is discussed in detail in the next section.
Patient preparation and anesthesia
The children coming to MRI are usually those who have undergone adenotonsillectomy but still have obstructive sleep apnea on postoperative polysomnogram and are not tolerating CPAP or want to consider surgical options. Because symptoms are only present during sleep, MRI is performed during anesthesia, akin to drug-induced sleep endoscopy. Obtaining ideal dynamic airway images during sleep MRI studies requires that the child be in a state mimicking physiological sleep as closely as possible, while maintaining spontaneous breathing via the native airway. Maintaining cardiorespiratory parameters such as oxygen saturation, respiratory rate and blood pressure within a safe range can be a challenge during dynamic airway evaluation . Historically, sedation has been performed under radiology’s structured sedation program with pentobarbital [1, 11]. However, it is optimal for anesthesia specialists to supervise all such studies. Initially, we used propofol for these studies, but it was recognized that this agent has dose-dependent respiratory depression  and the agent of choice became dexmedetomidine, which does not cause respiratory depression; use of dexmedetomidine is included in the procedure describe next .
The child is placed in neutral neck position with the Frankfort angle (angle from the inferior orbital rim to the auditory canal relative to horizontal) at 90°. Neck extension, shoulder rolls and other maneuvers are avoided to maintain as natural of a supine position as possible. A head strap is positioned under the head so that a mask can be used for CPAP if required to maintain airway patency . The child is put to sleep with sevoflurane in the induction room and intravenous access is secured. A bolus of dexmedetomidine (1 mcg/kg) is administered and then maintained with an infusion of 1 mcg/kg/h . Ketamine at 1 mg/kg might be added to the dexmedetomidine protocol, and this is used when a combination of drug-induced sleep endoscopy and cine MRI are performed together. The advantages of using this combination of drugs are beyond the scope of discussion for this review article; details of the sedation protocol are available in the references [15, 16].
Airway management during anesthesia
As stated, it is never desirable to have oral airways or laryngeal mask airways present when imaging the upper airway because it distorts and obscures the anatomy and physiology that is being imaged (Fig. 4). Usually an oral airway is placed during induction and is removed prior to commencement of imaging. We prefer CPAP via facemask to a nasal trumpet in order to maintain airway patency and oxygenation during cine MRI because it does not distort anatomy in an unnatural way, and pressure support is easily removed without arousing the child . There is a caveat to using a full face mask: On rare occasions the air pressure administered orally by a full facemask can push the tongue posteriorly and narrow the retroglossal airway. This occurs when there is no open path for air from the open mouth anteriorly to the posterior pharynx, usually caused by an enlarged tongue obstructing the oral passage of air . The next best option is a nasal trumpet, although this can distort the anatomy and is not visible on MRI.
When either CPAP or a nasal trumpet is needed to support the airway, it is best to perform all the imaging, including the cine sequences, with these in place. When image acquisition is completed, turn off the CPAP or remove the nasal trumpet and repeat the cine sequences to re-assess the native airway movement. CPAP is easiest to remove and children do not awaken with the change. Removing a nasal trumpet is more difficult and sometimes arouses the child.
Evaluation of airway motion
The normal airway changes during breathing because of physiological pressure changes created during the act of breathing and by the laws of fluid mechanics. For children with or without obstructive sleep apnea, the tendency for airway collapse during inspiration is countered by adequate neuromuscular action resulting in dilation; however, normal airway in a child without obstructive sleep apnea shows minimum change in caliber with breathing during sleep .
The midline sagittal cine is the plane that allows for visualization of the entire airway at the same time and is best for detecting multilevel airway collapse (Fig. 5). The nasopharynx is typically the narrowest portion of the upper airway  and in children adenoid regrowth is a common cause of narrowing. The retroglossal airway is often very dynamic and on sagittal images appears to collapse anterior to posterior. However, it needs to be interpreted in conjunction with the axial retroglossal and nasopharyngeal cine images to rule out a circumferential collapsing pattern. Motion can be deceiving on the sagittal images alone because volume averaging in the sagittal slice makes circumferential airway collapse appear to be occurring in an anterior–posterior plane only.
Glossoptosis collapse occurs with anterior and posterior motion of the tongue (Fig. 6). Glossoptosis is a dynamic process of the tongue in which the tongue moves posteriorly and then anteriorly. If the tongue contacts the posterior wall the surface adhesion might be more difficult for the dilator muscles to overcome. Glossoptosis is not necessarily correlated with worse obstructive sleep apnea when compared to hypopharyngeal collapse. Both can result in significant obstruction. Often the tongue is retro-positioned because of the size of the tongue or poor neuromuscular tone, but the static retro-position of the tongue is different from glossoptosis. This can be addressed surgically with hyoid suspension or a tongue suspension suture, which can be coupled with reduction in the tongue volume through posterior midline glossectomy.
Lateral wall collapse is the least common pattern of airway collapse  and seems to be related to the child’s ability to keep the tongue anteriorly positioned. In some cases there is an increase in the anteroposterior diameter while the lateral walls collapse. Seen with this is forward “thrusting” of the tongue in an effort to maintain a patent airway. This is typically addressed with lateral expansion pharyngoplasty or bony surgeries that expand the posterior airway such as genioglossal advancement, hyoid suspension and 2-jaw surgery.
Tongue thrusting is one of several compensatory motions (head bobbing, jaw thrusting and neck extension) that can be observed on the cine imaging. Head bobbing either up and down or back and forth in a nodding motion and neck motion often manifests as an accentuation of the extension of the neck. In children with Down syndrome, the tongue is often sticking out of the confines of the oral cavity between the teeth and the anterior portion is moving forward with each respiratory cycle in a thrusting motion. These motions are often very dramatic. A consistent relationship between these compensatory motions and severity of obstructive sleep apnea has not been proved. It could be that this is an effort to keep the airway open and the effectiveness of the movement is variable for each child.
Evaluation of anatomical structures
Anatomical evaluation is an important component of the MRI evaluation in obstructive sleep apnea in children who have persistent obstructive sleep apnea despite previous adenotonsillectomy because hypertrophied lingual tonsils and adenoid regrowth are all commonly seen in these children. Macroglossia, with the tongue enlarged relative to the size of the oral cavity, is also common . The T2-weighted images show regrowth of adenoids and enlarged lingual tonsils because of their bright signal as compared to a dark background. The isotropic proton-density 3-D images allow for precise reconstruction in any plane to show the relationship of anatomical narrowing to the thickness of these tissues. The palatine tonsils are generally absent because adenotonsillectomy is the first-line treatment for obstructive sleep apnea in this population, but if they are present, they can be assessed.
In the past, recurrence of the palatine tonsils was not a concern because the traditional tonsillectomy technique removed the entire encapsulated tonsil. However, more recently some surgeons are performing intracapsular partial tonsillotomies, and in these cases recurrence of the palatine tonsils can occur.
The soft palate and adenoid tissue can interact and cause airway obstruction at the level of the nasopharynx as the palate collapses against the adenoid tissue. A vertical-oriented, elongated palate extending past the mid tongue or close to the epiglottis is generally considered abnormal. Abnormal soft palates demonstrate mild increase in the T2 signal relative to the tongue and sometimes the soft palate is thickened. A number of surgical palatoplasty techniques (e.g., lateral expansion pharyngoplasty) used by otolaryngologists are intended to decrease retropalatal obstruction and improve airflow.
In addition, obstruction can occur from an enlarged or glossoptotic tongue. Frequently, the tongue pushes on the soft palate, either due to the tongue’s large size or its movement. The soft palate is then pushed up against the adenoid tissue or the nasopharyngeal ceiling, resulting in obstruction . In these cases, tongue reduction or forward displacement of the tongue through hyoid suspension, tongue suspension or bony surgery might be indicated.
Lingual tonsils lie on the posterior surface of the tongue and are typically less than 5 mm thick and often barely seen on T2-weighted MRI in normal children (Fig. 1). Enlargement might be related to removal of other tonsillar tissue, reflux and obesity [25, 26]. Gastroesophageal reflux is often exacerbated, with greater changes in intrathoracic pressure seen with obstructive sleep apnea from breathing against obstruction. When enlarged, the lingual tonsils fill the retroglossal airway, causing narrowing in the AP direction and filling the vallecular space, pushing the epiglottis posteriorly. Because of extreme motion at the base of the tongue in some children, the lingual tonsils can be difficult to evaluate if respiratory triggering is marginal, but we have found that T2-weighted imaging in the sagittal and axial planes with propeller technique are robust in reducing motion artifact. A recent meta-analysis by Kang et al.  looked at the effect of lingual tonsillectomy on obstructive sleep apnea in four pediatric studies and include 73 subjects. The meta-analysis showed a mean reduction of the apnea hypopnea index by 8.9, but an apnea hypopnea index of less than 5 was obtained in 51%, and an apnea hypopnea index less than 1 in only 17% .
Enlargement of the tongue is present in many of the children imaged for obstructive sleep apnea, especially in those with Down syndrome. Evaluation of tongue size is difficult, but helpful signs of significant enlargement include the posterior aspect of the tongue abutting and displacing the soft palate. Another helpful sign is the degree to which the tongue protrudes posteriorly over the line that extends from the anterior border of the trachea (Fig. 2). If the tongue is abnormally posteriorly positioned and the position is static, it is said to be retropositioned, and when there is AP motion as judged on the axial cine, it is glossoptosis. The tongue can be both retropositioned and glossoptotic.
Fatty infiltration of the tongue is often present on proton-density images when there is no fat suppression. This has been studied in adults and shown to be increased in obese children with obstructive sleep apnea compared to age- and weight-matched controls without obstructive sleep apnea. The same group of researchers showed decreased metabolism in these tongue tissues [28, 29]. This decrease in metabolism is probably related to increased visceral fat that is associated with metabolic syndrome and a prediabetic state. In general, pediatric patients are at risk of recurrent obstructive sleep apnea if they are obese.
Cine MR anesthesia-induced sleep studies are a useful adjunct to clinical assessment when evaluating children with persistent obstructive sleep apnea despite first-line treatment with adenotonsillectomy. In children who are intolerant of CPAP and entertaining surgical options, cine MR studies are especially helpful in identifying anatomical and physiological causes of airway collapse that can be addressed surgically.
This paper was supported by National Institutes of Health (NIH) grant RO1HL105206-01. The authors have indicated no financial conflicts of interest.
The project described was supported by the National Center for Advancing Translational Sciences of the NIH, under award number 5UL1TR001425-03. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Compliance with ethical standards
Conflicts of interest
- 4.Berry RB, Brooks R, Gamaldo DE et al (2017) The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specifications. Version 2.4. American Academy of Sleep Medicine, DarienGoogle Scholar
- 22.Bates A, Schuh A, Amine-Eddine G et al (2018) Assessing the relationship between movement and airflow in the upper airway using computational fluid dynamics with motion determined from magnetic resonance imaging. Clin Biomech. https://doi.org/10.1016/j.clinbiomech.2017.10.011
- 28.Kim C, Jackson N, Chawla S et al (2011) Quantification of tongue fat by novel MRI techniques and its relationship to BMI. Am J Respir Crit Care Med 183:A3678Google Scholar