Application of intelligent automatic segmentation and 3D reconstruction of inferior turbinate and maxillary sinus from computed tomography and analyze the relationship between volume and nasal lesion

Abstract

The nasal structure is closely related to nasal diseases. If an individual presents a discomfort in the nose for an extended period of time, the nasal tissue might have abnormal phenomena. Chronic sinusitis can cause nasal sinus shrinkage, especially the maxillary sinus, while allergic rhinitis can cause inferior turbinate pachynsis. Therefore, the volumes of the inferior turbinate and maxillary sinus are important indexes for otolaryngologists to judge nasal diseases. At present, the volumes of the inferior turbinate and maxillary sinus are estimated by radiologists manually contouring computed tomography (CT) images of the head and neck. The process consumes time and the results are not objective. This study aimed to propose an automatic recognition and volume calculation for the inferior turbinate and maxillary sinus by using image processing techniques. The back propagation neural network (BPNN) was used for automatic recognition of the inferior turbinate and maxillary sinus. Parametric template matching (PTM) and the sub-region similarity were used as feature inputs. The level set method (LSM) was applied to circle the contour of the inferior turbinate and maxillary sinus. The marching cubes algorithm was employed for 3D reconstruction and visualization. The volume information was obtained from the nonlinear regression curve. The accuracy and sensitivity of the automatic recognition results for inferior turbinate and maxillary sinus was 96.3 % and 95.1 %, respectively. The relationship between volume and nasal lesion has been analyzed as well.