YOLOv8-Based Segmentation and 3D Reconstruction of Alveolar Bone and Mandibular Canal in CBCT Images

Dental implant planning stands as a critical aspect of modern dentistry, with the segmentation of the Alveolar Bone (AB) and Mandibular Canal (MC) in Cone Beam Computed Tomography (CBCT) images serving as foundational steps in this process. However, current research faces a significant gap which is a lack of studies focusing on segmenting the AB and MC from CBCT slices that accurately follow the curve of the lower jaw. This is particularly crucial as segmentation in these slices is essential for precise dental implant planning. Current AB and MC segmentation research only addresses segmentation from coronal or axial slices in CBCT images. Notably, the current research also fails to address the crucial aspect of reconstructing the segmented slices into a comprehensive 3D view, which could significantly enhance the visualization of the segmentation data for dental implant planning. This study introduces a new approach utilizing YOLOv8 to segment AB and MC in CBCT slices that accurately conform to the lower jaw curve. Additionally, this study transforms the segmented slices into a unified 3D view by stacking each slice and employing linear interpolation to enhance the smoothness of the result. Our results indicate that Yolov8m yields the highest Dice Similarity Coefficient (DSC) of 90.15% and mAP of 99.5%, while YOLOv8l yields the highest Intersection over Union (IoU) of 85.98% for segmenting the AB and MC.