Abstract
Obstructive sleep apnea (OSA) is a common but underdiagnosed sleep disorder caused by recurrent collapse of the upper airway. While polysomnography provides definitive diagnosis, it is costly, time-consuming, and difficult to scale. Current screening tools are typically based on symptoms and demographic risk factors, whereas anatomical imaging could provide more direct information about the structures involved in airway obstruction. In this study, we examine whether magnetic resonance imaging (MRI)-derived three-dimensional tongue and pharynx surface morphology can be used for interpretable OSA screening. We segment both organs from T1-weighted MRI, reconstruct surface meshes, and classify OSA status using a Class Node Graph Attention Network (CGAT), which learns from mesh connectivity and produces surface-level attention maps. In a sex-balanced cohort of 880 diagnosed OSA cases and 880 rule-defined low-risk controls, the pretrained multi-organ CGAT achieved F1 = 0.763, recall = 0.888, and area under the receiver operating characteristic curve = 0.819. Multi-organ fusion improved performance over tongue-only and pharynx-only models, indicating that the two structures provide complementary shape information. Attention maps localized predictive evidence to posterior tongue and pharyngeal wall regions, aligning with known OSA-related airway narrowing. These findings suggest that MRI-derived upper-airway morphology may enable radiation-free, anatomically interpretable OSA screening, but clinical validation with apnea-hypopnea index measurements remains necessary.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ShapeMI_036.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=WCFtk0mffX
BibTex
@InProceedings{MokAry_Interpretable_MICCAISAT2026,
author = { Mokhtari Saghafi, Aryan AND Devine, Jay AND Pauwels, Kaat AND Claessens, Nina AND Claes, Peter},
title = { { Interpretable Obstructive Sleep Apnea Screening from MRI-Derived Upper-Airway Surface Morphology with Class Node Graph Attention Networks } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17259},
month = {pending},
page = {pending}
}