Three-Dimensional Image Segmentation of Upper Airway by Cone Beam CT: A Review of Literature

Abstract

The aim of this paper is to review the different approaches of three-dimensional image segmentation using cone-beam computed tomography (CBCT), with a focus on the human upper airway. Literature reviews in the dental field have been included, relating to the use of CBCT to assess the upper airway using image segmentation. Obstruction of the upper airway often modifies normal breathing, which can have a noticeable impact on the typical development of craniofacial structures. CBCT is a modality that allows for the improved understanding of airway anatomy, pathology and upper airway analysis. It is more accurate, efficient and has a relatively less radiation dose compared to multi-detector CT. Three-dimensional (3D) models of the upper airway that have been segmented and designed from CBCT images can be used to visualize and analyze treatment efficiency in subjects with breathing or obstruction disorders. The accuracy of these 3D morpho-functional analytical models is essential for improving diagnosis, treatment planning, and assessing treatment outcomes of the upper airway. Therefore, the purpose of the present review is to discuss the different methods for upper airway segmentation using CBCT to achieve accurate modeling.

Artificial Intelligence as an Aid in CBCT Airway Analysis: A Systematic Review

“…The first and most important stage in assessing the upper airway volume is segmentation, which allows the airway space to be distinguished from the rest of the scan and can then be visualized and quantified in three dimensions. Various upper airway segmentation methods and algorithms that are either manual, semiautomatic, or automatic in nature have been developed over the past ten years [31]. Even while hand segmentation is the gold standard and provides the most precise replication of the anatomical structure, it is labor-and time-intensive.…”

Guías de uso de tomografía computarizada de haz cónico en ortodoncia: revisión narrativa

“…(10,16) Esto puede deberse al limitado valor diagnóstico que ha mostrado en esta área. (25,26) Además, no existe un protocolo de posicionamiento único para obtener las imágenes de la vía aérea antes y durante un tratamiento. (25,26) Por otra parte, el volumen de la vía aérea cambia según las diferentes fases respiratorias, el posicionamiento de la cabeza y la lengua, e inclinación craneocervical.…”

“…(25,26) Además, no existe un protocolo de posicionamiento único para obtener las imágenes de la vía aérea antes y durante un tratamiento. (25,26) Por otra parte, el volumen de la vía aérea cambia según las diferentes fases respiratorias, el posicionamiento de la cabeza y la lengua, e inclinación craneocervical. (27) Por lo tanto, se requiere un protocolo de posicionamiento estandarizado, para reducir las alteraciones volumétricas y craneocervicales durante la toma del examen.…”

“…Debido a que la mayoría de los estudios tienen un bajo nivel de evidencia, es difícil generar indicaciones sobre la evaluación 3D de la vía aérea. (25,26) Se necesitan más estudios para establecer el impacto de la CBCT en la evaluación 3D de la vía aérea.…”

Automatic segmentation of the pharyngeal airway space with convolutional neural network

“…The first and the most vital step for analyzing PAS volume consists of segmentation which enables the delineation and separation of the airway space from the rest of the scan, thereby, allowing 3D visualization and quantification. Over the past decade, various upper airway segmentation tools and algorithms have been developed which are either manual, semiautomatic or automatic in nature [11]. Although manual segmentation does offer the most accurate replication of the anatomical structure and is considered as a gold standard, nevertheless, it is time-consuming and labor-intensive.…”