Application of Mask-RCNN and Mimics in Maxillary Sinus Modeling
Objective:To compare the application of Mask RCNN deep learning model and Mimics 3D software in maxillary sinus modeling.Methods:Mask-RCNN and Mimics were applied to reconstruct the maxillary sinus and measure the volume of maxillary sinus from conical beam CT images in 50 patients included.The reconstruction effects of the two methods were compared,and the volume of the maxillary sinus was analyzed.Results:In the process of modeling the maxillary sinus,using Mask-RCNN for image segmentation,post-processing,and recon-struction only took more than 30 seconds,and using Mimics for manual threshold segmentation and reconstruction of maxillary sinus images for each patient took about tens of minutes.There was no significant difference in the vol-ume of the maxillary sinus measured between two methods(P＞0.05).Conclusion:The Mask RCNN deep learning algorithm is superior to Mimics and can reconstruct the maxillary sinus more quickly and accurately,reflecting the auxiliary diagnostic value of artificial intelligence in the field of oral and maxillofacial medical imaging.
万方数据
维普
