Introduction

Obstructive sleep apnea (OSA) is a common sleep-related breathing disorder caused by upper airway collapse, with a global prevalence ~1 billion and a significant healthcare burden¹. OSA is diagnosed by polysomnography, and anatomical imaging including proton MRI can be used to identify sites of airway collapse for treatment planning. However, there is a lack of robust methods for evaluating the relationship between airway anatomy and airflow; for example, regions of fast-moving airflow indicate obstructed sections of the upper airway, which may be candidates for surgical interventions.
Phase contrast (PC) MRI offers a means to measure fluid flow non-invasively in-vivo, and is commonly used in cardiovascular medicine². Hyperpolarized (HP) gases have shown utility as inhaled contrast agents for PC MRI of the airways^3-5, and we recently reported a concordance between PC MRI and computational fluid dynamics (CFD) simulations of upper airway flow in healthy volunteers⁶. To date, in-vivo HP gas PC MRI studies have been limited to healthy volunteers^{7, 8} or small animals⁵.
Here, we hypothesize that HP ¹²⁹Xe PC MRI is feasible in pediatric patients with OSA, and can aid the assessment of respiratory flow in the upper airway by identifying obstructions.

Methods

3 patients (mean age=15.6years, obstructive apnea-hypopnea index=11.4) were enrolled in this study, which was approved by the Cincinnati Children’s Hospital IRB.
PC MRI was performed using a 3T Philips clinical MRI scanner and flexible ¹²⁹Xe radiofrequency coil, during a clinically-indicated MRI performed for OSA. Previously, we reported ¹²⁹Xe PC MRI of the upper airways in non-sedated healthy volunteers during slow inhalation of 1L of ¹²⁹Xe over a period of 10—20s, during which a series of PC images were obtained⁶. In the present work, patients with OSA were clinically sedated with dexmedetomidine, which approximates sleep stage-2, and ¹²⁹Xe gas (1L, polarized to ~30%) was delivered from a bag to the patient’s anesthesia mask during tidal breathing and was removed after two inhalation-exhalation cycles.
2D sagittal PC MRI data were acquired continuously during this period, with acquisition parameters; spiral readout, spatial resolution: 2.5mm², slice thickness: 10mm, venc: 200cm/s, velocity encoding in either one (foot-head) or three spatial directions (right-left, anterior-posterior, foot-head); acquisition time for a single slice: ~0.5 or 1s, respectively.
Flow at the mask was measured with an MRI-compatible pneumotachograph; patient heart rate and SpO₂ were measured throughout.

Results

The MRI procedure was successful, with no adverse events. SpO₂ did not noticeably drop at any time during or after the introduction of ¹²⁹Xe gas.
Velocity maps of respiratory airflow over two breaths in the upper airway and trachea along the foot-head direction are shown in Figure 1. In this example subject, three images captured inspiratory airflow, two captured expiration, and one captured the transition. During inhalation, a high-speed jet formed in the nasopharynx, which qualitatively corresponds to a region of airway narrowing at the tip of the soft palate on ¹H anatomical MRI. High velocity flow is not visible in the same region during exhalation, demonstrating that obstruction occurs only during inhalation. A jet formed through the glottis during both inhalation and exhalation, corresponding to the region inferior to airway narrowing at the epiglottis on ¹H anatomical MRI. Results obtained from 3-directional PC MRI in the same subject (Figure 2) illustrate that despite the 2-fold lower temporal resolution compared with unidirectional PC MRI, respiratory airflow during inspiration and expiration was captured by at least one distinct dynamic image. Although the most pronounced jets are observed in the (dominant) foot-head direction, anterior-posterior flow appears to show additional features that may be informative in patients who exhibit more turbulent flow patterns.
The approximate mean SNR in the nasopharynx for the 3-directional scan was 21.7±4.8 (inspiration phase) and 10.0±5.5 (expiration), and for the FH-only scan was 23.1±8.3 (inspiration) and 8.7±5.9 (expiration). The inspiratory/expiratory patterns of airflow observed in PC MRI qualitatively matched the breathing cycle measured by the pneumotachograph (Figure 3).

Discussion

Low SNR of ¹²⁹Xe PC MRI data, particularly in the expiration phase, constrains the accuracy of airflow velocity measurements⁹. Whilst 1L ¹²⁹Xe was connected to the gas delivery apparatus, the actual volume of gas that enters the airways during two breathing cycles is likely much less. The addition of a preparatory gas “wash-in” phase may boost SNR. In addition, these experiments were performed using a flexible chest coil with suboptimal efficiency when used for upper airway measurements; a localized receiver array proximal to the neck may provide SNR benefits.
Though spiral trajectories are inherently robust to motion, bulk motion of the airway during acquisition can cause motion artefacts and difficulties in isolating the airway signal. In patients with more complicated bulk airway motion, the use of image acceleration techniques⁷, motion-compensated reconstruction strategies, in combination with ¹H-MRI-guided image registration may be needed.

Conclusion

We have demonstrated that ¹²⁹Xe PC MRI can be performed safely to map regional airway velocity in patients with OSA under sedation, with sufficient sensitivity to identify regions of obstruction at different phases of the breathing cycle.
Combined with proton MRI of airway anatomy and CFD simulations, this may lead to further understanding of disease severity and local function, allowing the development of personalized treatment plans.