Toward Clinically Applicable 3-Dimensional Tooth Segmentation via Deep Learning

Abstract

Digital dentistry plays a pivotal role in dental health care. A critical step in many digital dental systems is to accurately delineate individual teeth and the gingiva in the 3-dimension intraoral scanned mesh data. However, previous state-of-the-art methods are either time-consuming or error prone, hence hindering their clinical applicability. This article presents an accurate, efficient, and fully automated deep learning model trained on a data set of 4,000 intraoral scanned data annotated by experienced human experts. On a holdout data set of 200 scans, our model achieves a per-face accuracy, average-area accuracy, and area under the receiver operating characteristic curve of 96.94%, 98.26%, and 0.9991, respectively, significantly outperforming the state-of-the-art baselines. In addition, our model takes only about 24 s to generate segmentation outputs, as opposed to >5 min by the baseline and 15 min by human experts. A clinical performance test of 500 patients with malocclusion and/or abnormal teeth shows that 96.9% of the segmentations are satisfactory for clinical applications, 2.9% automatically trigger alarms for human improvement, and only 0.2% of them need rework. Our research demonstrates the potential for deep learning to improve the efficacy and efficiency of dental treatment and digital dentistry.

AI-enabled Automatic Multimodal Fusion of Cone-Beam CT and Intraoral Scans for Intelligent 3D Tooth-Bone Reconstruction and Clinical Applications

“…The IOSNet achieved a mIoU of 93.07% and 95.70% on the maxillary and mandible IOS scans, 1.77% and 3.52% higher than best-performing baselines. Given that DCNet was already able to generate clinically applicable results for most cases 18 , such a significant improvement could corroborate better performance for the IOSNet. The performance of IOSNet was subsequently demonstrated by the real-world clinical validation with our DDMA framework.…”

“…The IOS segmentation module, termed IOSNet, aimed to generate predictions for each face in the IOS meshes. It first transformed the input mesh into point clouds, then processed with the modified DCNet architecture 32,18 . The previous DCNet had limitations in precise boundary segmentation or generalization to complicated anatomies, like crowded teeth and hyperdontia.…”

“…. We evaluated the performance of IOSNet over the same test set as our pioneering work, DC-Net 18 . The test set consisted of 200 meshes from 100 patients.…”

“…The goal of the deep learning-based segmentation system is to perform point-wise classification for the point cloud obtained in the data preprocessing step. The data preprocessing follows the DC-Net baseline to convert the original IOS mesh to a point cloud with 15-dimensional features as in the DC-Net 18 . Deep learning segmentation.…”

“…Therefore, the IOS models could provide the accuracy and occlusal information needed in CBCT for clinical applications. Recently, our group and others have developed methods for automatic segmentation of the IOS models using deep learning with high accuracy [18][19][20][21] . Herein, the automatic fusion of both IOS and CBCT segmentations could provide crown-root-bone structures with the accuracy needed in many dental applications.…”