GENERALIZABILITY OF CLOUD-BASED AI SOFTWARE FOR ANTERIOR TOOTH SEGMENTATION IN MULTICENTER CBCT DATASETS: AN EXTERNAL VALIDATION STUDY.

A novel fluoride tray-assisted workflow for artificial intelligence-driven automated maxillary gingival segmentation on cone beam CT

Abstract Objectives To clinically validate an artificial intelligence (AI)-based tool for automated maxillary gingival segmentation of marginal and supracrestal gingiva on cone beam CT (CBCT) scans using a novel soft tissue separation technique with a fluoride tray. Methods A validation set of 35 CBCT scans, covering maxilla, acquired with fluoride trays was processed using a cloud-based AI platform (Relu, Leuven, Belgium) for gingival segmentation. The resulting models were refined by an expert using 3-dimensional (3D) mesh-processing software and compared with the original AI outputs to assess accuracy. Additionally, 6 CBCT scans were manually segmented, using intraoral scans as reference, and compared with the AI model. Surface-based and voxel-wise analyses, color-coded maps, consistency, and time-efficiency were evaluated. Wilcoxon signed-rank test was used to assess time differences among methods. Results Artificial intelligence vs expert refinement showed strong agreement with high overlap (medianDSC ≥ 96%) and minor surface deviations (medianMSD∼0.00 mm). Minor differences were found between anterior and posterior regions (medianΔDSC = 1%, ΔMSD∼0.00 mm). Artificial intelligence vs manual segmentations showed median dice similarity coefficient (DSC) of 82% and small median MSD of 0.15 mm. Labial/buccal area showed the highest surface deviations from color-coded maps. Bland-Altman plots showed low intra- and inter-operator consistency in time, while AI showed excellent consistency. Artificial intelligence demonstrated significantly faster segmentation, achieving 4x faster with expert refinement and 20x faster with manual approach. Conclusion The use of fluoride tray facilitates separation of oral soft tissues on CBCT scans, enabling accurate and time-efficient AI-driven segmentation of maxillary marginal and supracrestal gingiva. This supports integration into digital workflows and more efficient treatment planning. Advances in knowledge The use of a fluoride tray for soft tissue separation during CBCT scans facilitates gingival visualization while maintaining patient comfort. The resulting 3D gingival models from AI-based segmentation can be integrated with automatically segmented dentomaxillofacial structures, enhancing clinical visualization of oral soft and hard tissues and facilitating diagnosis and treatment planning, including periodontal evaluation, implant planning, prosthodontics, and orthodontics.

Diagnostic Performance of Deep Learning for Automated Mandibular Canal Segmentation on CBCT Images: A Systematic Review and Meta-Analysis.

INTRODUCTION Accurate localization of the mandibular canal in Cone-Beam Computed Tomography (CBCT) images is critical for preventing iatrogenic nerve injury during maxillofacial surgery and dental implant procedures. This systematic review and meta-analysis aimed to evaluate the diagnostic performance, anatomical localization accuracy, and time efficiency of deep learning-based artificial intelligence (AI) systems in automated mandibular canal segmentation compared to traditional manual expert annotations. MATERIALS AND METHODS A comprehensive literature search was conducted across PubMed, Scopus, Web of Science, IEEE Xplore, and Embase databases in accordance with PRISMA guidelines. Studies evaluating the performance of AI models for mandibular canal detection on CBCT scans using expert annotations as the reference standard were included. The primary outcome measure was the Dice Similarity Coefficient (DSC), while secondary outcomes included Average Symmetric Surface Distance (ASSD) and processing time. Statistical analyses were performed using a random-effects model. RESULTS A total of 38 unique studies comprising over 8,420 CBCT volumes were included in the quantitative synthesis. The pooled DSC for AI-driven segmentation was calculated as 0.82 (95% CI: 0.79-0.85). Subgroup analyses revealed that transformer-based architectures (DSC: 0.89) demonstrated significantly superior performance compared to traditional convolutional neural networks (CNNs). The pooled ASSD exhibited a high anatomical accuracy of 0.42 mm (95% CI: 0.38-0.47), which is close to voxel dimensions. Furthermore, the autonomous segmentation process was completed in an average of 32 seconds, whereas manual expert annotation took 600 seconds (p < 0.001), confirming an 18.7-fold timesaving in the clinical workflow. DISCUSSION Deep learning algorithms provide highly accurate, reproducible, and time-efficient results at a human-expert level in the automated segmentation of the mandibular canal on CBCT images. The integration of these AI systems into clinical protocols has the potential to enhance surgical safety and standardize preoperative planning processes in dental implantology.

Two-Stage 3D U-Net-Based Segmentation Of Maxillofacial Structures From CBCT Scans For Dental Implant Planning

This work represents a 3D segmentation framework for automatic landmark delineation in Cone Beam Computed Tomography to aid dental implant planning. The proposed method aimed to segment important anatomical details like bone, nerve canals, and teeth using the Bone segmentation algorithm, Deep Neural Network Based Minimal Medical segmentation, and Tooth segmentation method, respectively. The bone segmentation algorithm utilizes two-stage 3D U-Net architecture where coarse global segmentation combines with fine segmentation for accurate detection of anatomical structures like tooth, alveolar ridge, maxillary sinus and nasal cavity. Preprocessing of CBCT images is done preceding to segmentation to remove poisonous noise using PNR algorithm, Thermal noise using TFR algorithm and ring artifacts with RAR algorithm. The regions of interest identified in segmentation phase leads to exact measurements of important anatomical structures that further assists in automation of medical report generation system. The assessment of segmentation algorithms is done using parameters such as DICE-Score and IoU. Our proposed system achieved 0.96, 0.97, 0.92, 0.96 and 0.99 DICE-Score values for mandible, maxilla, nerve canal, teeth and background respectively as well as 0.93, 0.94, 0.86, 0.93 and 0.99 IoU score values for mandible, maxilla, nerve canal, teeth and background respectively.

A Scalable, Annotation-Free Pipeline for Automated CT Pelvimetry: A Validation Study with Landmark Uncertainty Analysis and Clinical Correlation.

Deep learning for automated instance segmentation of residual roots in panoramic radiographs using mask R-CNN: A retrospective diagnostic accuracy study

Background: This study aimed to evaluate the efficacy of Mask regions with convolutional neural network (R-CNN) for the automated detection and segmentation of residual dental roots in panoramic radiographs and to compare its diagnostic performance against the semantic segmentation benchmark, U-shaped network (U-Net). Methods: A retrospective dataset comprising 224 patients with 505 annotated residual roots was utilized. Image preprocessing involved adaptive contrast enhancement using contrast limited adaptive histogram equalization in the Commission Internationale de l’Éclairage lab color space to improve root-to-bone definition. A Mask R-CNN model utilizing a ResNet-50 backbone and Feature Pyramid Network was trained using K-fold cross-validation. Performance was compared to U-Net based on sensitivity, specificity, accuracy, dice similarity coefficient, and receiver operating characteristic analysis. Result: The Mask R-CNN model significantly outperformed U-Net across all evaluated metrics. It achieved an accuracy of 98.67% and a dice similarity coefficient of 91.34%. Most notably, the model demonstrated a sensitivity of 91.16%, presenting a marked improvement over U-Net (78.54%), while maintaining a specificity of 99.12%. The area under the curve was calculated at 0.9599, indicating superior discriminative capability. Conclusion: Mask R-CNN provides a robust solution for identifying residual roots, effectively addressing challenges related to low contrast and anatomical noise. By combining high sensitivity with high specificity, the system significantly reduces false negatives without causing alert fatigue, thereby serving as a reliable automated assistant for enhancing surgical safety and planning efficiency.

Automatic Annotation of Cephalometric Landmarks using a Two-Stage Multi-Regional Context-Enhanced Framework in Cone-Beam Computed Tomography Volumes.

OBJECTIVES In this retrospective study, we aimed to develop an incomplete-observation-aware framework for automatic three-dimensional cephalometric landmark localization in cone-beam computed tomography (CBCT) volumes, capable of explicitly modelling landmark visibility, reconstructing missing anatomical context, and effectively leveraging both fully and partially annotated clinical data. METHODS An incomplete-observation-aware framework was developed for CBCT cephalometric landmark localization. The framework explicitly models landmark visibility under limited field-of-view conditions and incorporates latent anatomical completion within a two-stage, coarse-to-fine localization strategy. A total of 144 CBCT scans obtained for routine orthodontic diagnosis and treatment planning, including 90 fully annotated and 54 partially annotated, were retrospectively collected, with 115 cases for training and 29 for testing. Performance was evaluated using the mean radial error (MRE) and successful detection rate (SDR). RESULTS The proposed framework achieved an MRE of 1.01 mm ± 0.70 mm on the complete-landmark test set and 1.01 mm ± 0.72 mm on the mixed-landmark test set. SDRs exceeded 91% within 2 mm and reached 100% within 6 mm for both datasets. The comparable performance observed across the complete and mixed test sets demonstrates the robustness of the proposed framework in handling incomplete observations and partially annotated data. CONCLUSIONS The proposed framework achieved robust and accurate performance for automatic three-dimensional (3D) cephalometric landmark localization on CBCT volumes by modelling landmark visibility and incorporating latent anatomical completion, enabling reliable localization under incomplete observations. CLINICAL RELEVANCE The proposed framework enables automated 3D cephalometric landmark localization in CBCT volumes and is robust to incomplete field-of-view and missing landmarks through visibility-aware modelling and latent anatomical reconstruction.