Automated Detection of Missing Tooth Regions in CBCT Slices for Dental Implant Planning Using CNN with Transfer Learning

Detecting the missing tooth region in Cone Beam Computed Tomography (CBCT) slices is crucial for dentists when planning dental implant placement. It allows dentists to accurately identify regions where tooth is missing and assess the suitability of these regions for dental implant placement. However, manually detecting CBCT slices to determine these areas can be time-consuming for dentists. Previous studies have primarily focused on panoramic radiographs or 3D CBCT volumes to detect missing tooth, but none have classified missing tooth in CBCT slices that follow the jaw curve, which is crucial for dental implant planning. This study proposes a new approach that leverages CNN with transfer learning using well-known architectures: InceptionV3, VGG16, MobileNetV2, ResNet50, and ResNet101 to detect missing tooth regions for dental implant placement in Cone Beam Computed Tomography (CBCT) slices that follow the jaw curve. Additionally, histogram equalization enhances the CBCT slices before the classification process. K-fold cross-validation with k=5 is employed to evaluate the performance of the proposed method. The results indicate that the ResNet101 architecture achieves the highest average F1 score of 98.2%. Automated detection of missing tooth regions can significantly reduce the time and effort required to plan dental implants, allowing dentists to focus more on patient care.