Novel dynamic upper airway MRI methods to characterize obstructive sleep apnea

Obstructive sleep apnea (OSA) is characterized by the dynamic obstruction of the upper airway during sleep. If left untreated, OSA can lead to severe health complications, including arrhythmia and stroke. The first-line treatment for OSA is continuous positive airway pressure (CPAP) therapy. However, 30-70\% of OSA patients find it challenging to adhere to CPAP due to factors such as claustrophobia, pressure discomfort, and machine noise. Alternative therapies are available but rely on appropriate patient selection, and the current standard for patient selection, drug-induced sleep endoscopy (DISE), has several limitations. DISE does not represent natural sleep due to the use of anesthesia and lacks the ability to simultaneously image the entire upper airway (from the nasopharynx to the hypopharynx) and the surrounding soft tissues (e.g., tongue, soft palate) that contribute to airway collapse. In contrast, magnetic resonance imaging (MRI) is a non-invasive modality that offers excellent contrast for visualizing muscles and soft tissue structures, making it a promising tool for assessing upper airway dynamics during normal and impaired breathing. However, significant technical challenges hinder the routine use of MRI in evaluating airway collapse during OSA. These challenges include: a) the absence of MR receiver coil hardware capable of comfortably imaging patients in various sleep postures (necessary to observe posture-dependent airway obstruction), b) the difficulty in achieving both adequate spatial and temporal resolution---specifically, at least 2.0 mm\(^2\) spatial resolution and 200 ms per frame temporal resolution (required to reliably capture dynamic changes in the airway's cross-sectional area), and c) the challenge of obtaining full upper airway coverage from the nasopharynx to the hypopharynx (necessary for simultaneous multi-site airway evaluation). To address the first challenge in MRI-based OSA assessment, this thesis presents the development of a novel custom coil for efficient parallel imaging of the upper airway. This work includes the design of a 16-channel airway coil with elements strategically positioned to provide high sensitivity in all pertinent OSA-related regions (e.g., tongue, soft palate, pharynx, glottis, epiglottis, vocal cords). The coil design is compact and segmented into three parts, allowing for adaptable orientation to accommodate patients with varying head and neck sizes. The remaining challenges are addressed by developing an information-efficient, motion-robust, data-driven 2D multi-site dynamic MRI scheme to quantify airway collapse in OSA, focusing on regions from the velopharynx to the hypopharynx. Finally, this thesis demonstrates a C++-based framework with the potential to bring the developed offline reconstruction scheme into a live environment for online reconstruction, thereby streamlining the sleep scan study pipeline. The immediate application of these methods together in clinical trials may prove to be superior to DISE for the pre-surgical evaluation of non-CPAP therapies.