Abstract
BACKGROUND AND OBJECTIVE: Three-dimensional (3D) reconstruction of head and neck tissues has extensive clinical applications, but due to the complexity and variability of tissue structure, there is still a lack of a complete scheme to reconstruct the head and neck tissues. This study aims to establish a tissue-specified multi-directional cross-sectional image sequence construction method to capture diverse tissue contour information. METHODS: The image sequences that are most conducive to acquiring the boundary contours of the target tissue are constructed from 3D MRI images of the head and neck in a non-traditional way based on the characteristics of each target tissue, and an effective registration strategy is used to integrate the boundaries of the target tissue segmented from multiple image sequences. The NURBS (Non-Uniform Rational B-Splines) surface modeling method is used to construct the 3D structure of the head and neck based on the segmented tissue boundaries, and then the constructed structure is used to build a fluid-structure interaction model to simulate airway collapse. RESULTS: The multi-directional cross-sectional image sequences of head and neck tissues were reconstructed, which successfully supplemented the missing boundary information in unidirectional image sequences commonly used in anatomical reconstructions. The boundaries of the tongue and soft palate were obtained from three corresponding sequential images respectively, and nonlinear registration methods were developed to match the intersections of the target tissue boundaries segmented from different image sequences. The complete 3D head and neck structure, including the surrounding tissues of the upper airway, was accurately reconstructed, and then directly converted into a finite element model through a meshing procedure. The head and neck numerical models successfully simulate airway collapse in both the obstructive sleep apnea patient and the normal subject, providing detailed information on soft tissue deformation and predicting the values of the airway critical closing pressure. CONCLUSIONS: A complete 3D reconstruction scheme from multi-directional image sequence construction to nonlinear boundary registration and NURBS surface generation is established. The constructed model can accurately reflect the characteristics of real anatomical structure, and can be directly used for complex numerical simulations of upper airway collapse.