SUMMARY An estimated 30-70% of obstructive sleep apnea (OSA) patients are intolerant to first-line continuous positive airway pressure (CPAP) therapy for a variety of reasons. A variety of alternatives to CPAP exist, including the 2014 FDA approved hypoglossal nerve stimulation (HNS) therapy. The current clinical practice to screen OSA patients for HNS surgery is to perform a drug induced sleep endoscopy (DISE) procedure to assess airway collapse. However, DISE is challenged by several limitations including potential alteration of natural airway collapse due to the inserted endoscope, the lack of imaging surrounding soft-tissue, the lack of simultaneously imaging the full 3D extent of the upper airway, and the purely qualitative interpretation of airway collapse. Despite these challenges, it is mainly used because there is no better alternative. In this proposal, we propose to address this gap by systematically developing a novel 3D dynamic MRI (3D-DMRI) scheme capable of quantitating airway collapse during sleep. We propose a novel scheme that synergistically combine (a) parallel imaging via dedicated airway coils, (b) motion robust 3D stack of spiral encoding, (c) learning based spatio- temporal regularization, and (e) quantitative analysis to extract imaging phenotypes to characterize 3D airway collapse. In Aim 1, we will develop the scheme and systematically validate it on a cohort of OSA patients being screened for HNS therapy. In Aim 2, we will develop a deep transfer learning based segmentation scheme to extract 3D-DMRI based phenotypes that characterize spatio-temporal patterns of airway collapse. In Aim 3, we will assess the clinical utility of our scheme to characterize the pattern of airway shaping and collapse both in the awake and sleep state, and will evaluate it against existing DISE clinical protocol. In aim 4, we will determine if the 3D-DMRI derived imaging-based biomarkers are predictive of clinical outcomes of HNS therapy. This study will leverage the principal investigator’s expertise in rapid upper-airway MRI, along with collaborators’ interdisciplinary expertise in otorhinolaryngology, clinical sleep medicine, image analysis. Successful completion of this study will showcase the feasibility of MRI-based “virtual endoscopy” and immediately lead to several clinical trials for OSA. Specifically, we anticipate having demonstrated 3D-DMRI to be an effective alternative to DISE for determining the eligibility of patients with OSA for HNS, and show preliminary evidence if these biomarkers are predictive for HNS treatment success.