Introduction:

The human upper airway is vital for physiological functions like respiration and phonation, making its study crucial for diagnosing conditions such as obstructive sleep apnea [1], aiding surgical planning, and informing speech therapy [2]. Magnetic Resonance Imaging (MRI) is the preferred, non-invasive method for this study [3], providing structural and dynamic soft-tissue detail. The utility of 3D static MRI in generating Finite Element Method (FEM) models is clear [4], yet its application is hampered by the need for manual segmentation, a process that is notoriously time consuming and costly, requiring significant expert hours per data volume. While deep learning offers a potential solution for segmentation, its effectiveness is limited by the scarcity of large, labeled, domain-specific datasets, despite advances like transfer learning. This critical gap necessitates expert annotated data, yet the lack of robust automated tools means creating these datasets remains a labor intensive-endeavor.

Method:

Our proposed work directly addresses these challenges by making two significant contributions to the speech MRI community. Firstly, it provides a unique, publicly available dataset comprising volumetric MRI data from 20 English speakers, including six static voiced images per subject, their expert segmentations, and 3D printable vocal tracts. This dataset is foundational for developing and validating new state-of-the-art segmentation algorithms. Secondly, it introduces a no-code segmentation tool built using Slicer, Monal, Label, and Docker, designed to enable non-technical experts to generate preliminary segmentations, thus streamlining the dataset creation process. By making both the code and model weights open source, this research offers accessible resources that will significantly accelerate the advancement of airway analysis and facilitate the creation of extensive datasets in the field.

Results: Segmentations are ongoing, and will result in a unique, publicly available dataset.

REFERENCES
1. Daley, J. T. & Schwab, R. J. Imaging the airway in obstructive sleep apnea. in Neuroimaging of Sleep and Sleep Disorders (eds. Nofzinger, E., Thorpy, M. J. & Maquet, P.) 256–263 (Cambridge University Press, Cambridge, 2013). doi:DOI: 10.1017/CBO9781139088268.035.
2. Wong, A.-M. et al. Examining the impact of multilevel upper airway surgery on the obstructive sleep apnoea endotypes and their utility in predicting surgical outcomes. Respirology 27, 890–899 (2022).
3. Darquenne, C. et al. Upper airway dynamic imaging during tidal breathing in awake and asleep subjects with obstructive sleep apnea and healthy controls. Physiol Rep 6, e13711 (2018).
4. Xu, C., Brennick, M. J., Dougherty, L. & Wootton, D. M. Modeling upper airway collapse by a finite element model with regional tissue properties. Med Eng Phys 31, 1343–8 (2009).