Automatic 3D Quantification Of Root Resorption Using Deep Learning

Orthodontically induced root resorption (OIRR) is a common complication in orthodontic treatment, and this study explores the use of deep learning for accurate, automated quantification and localization of OIRR from CBCT images, addressing the limitations of traditional, time-consuming, and subjective manual assessments. This retrospective, cross-sectional study analyzed 4534 teeth from 105 patients aged 12–35 years. CBCT images were preprocessed for automatic tooth segmentation and converted into point clouds. A Dynamic Graph Convolutional Neural Network (DGCNN) was used to segment crowns and roots. The study was approved by the Ethics Committee of the Affiliated Hospital of Stomatology, School of Stomatology, Zhejiang University School of Medicine, approval number (2024-001[R]). The intraclass correlation coefficient was employed to validate the consistency between the automatic model and manual measurements. The proposed method strongly correlated with manual measurements in terms of root volume and OIRR severity assessment. The intraclass correlation coefficient values for average volume measurements at each tooth position exceeded 0.95 (P<0.001), with the accuracy of different OIRR severity classifications surpassing 0.8. The proposed methodology provides automatic and reliable tools for OIRR assessment, offering potential improvements in orthodontic treatment planning and monitoring.