Detection of Missing Tooth Regions Using Deep Learning in Panoramic Radiographs for Dental Implant Planning

The advances in dental radiology, particularly the utilization of panoramic radiographs, have significantly enhanced the precision of dental implant planning. This study introduces a novel deep learning-based approach for detecting missing tooth regions in panoramic radiographs, leveraging a region-based Convolution Neural Network (Mask R-CNN) with a Residual Neural Network (ResNet-101) to enhance the extraction of features from input data, such as the backbone for tooth segmentation and numbering. By integrating a heuristic algorithm, the proposed method improves detection accuracy and addresses common challenges such as multiple numbering errors and misalignments. The model was evaluated using a robust dataset, demonstrating superior performance metrics, including a precision of 0.9566, a recall of 0.9635, and a mean Average Precision (mAP) of 0.9241, compared to conventional methods. The results affirm the potential of this automated system to streamline dental implant planning, reduce clinician workload, and support advanced diagnostic and educational tools.