Deep Learning for Automatic Segmentation of Bone Sequestra on Orthopantomographs: A UNet-Based Approach
Deep Learning for Automatic Segmentation of Bone Sequestra on Orthopantomographs: A UNet-Based Approach
This study developed and evaluated a UNet-based deep learning model for automatic segmentation of bone sequestra on orthopantomographs (OPGs), with the goal of improving diagnostic efficiency and reducing inter-observer variability in osteomyelitis, osteoradionecrosis, and medication-related osteonecrosis of the jaw. A total of 120 anonymized OPGs with 134 annotated sequestra were used for training. Images were preprocessed with min-max normalization and augmented to enhance robustness. Manual expert annotations served as ground truth. A UNet architecture, optimized with Dice loss and the AdamW optimizer, was trained for 300 epochs. Performance was assessed using Dice coefficient, Intersection over Union (IoU), precision, recall, F1-score, and accuracy. ROC analysis and confusion matrix evaluations were performed. Agreement with clinicians was quantified using the Intraclass Correlation Coefficient (ICC). The model achieved a best-checkpoint validation Dice of 0.79 and IoU of 0.93. On the test set, performance included a Dice of 0.79, IoU of 0.74, recall of 0.81, precision of 0.81, and F1-score of 0.81. ROC analysis showed balanced discriminative performance with an AUC of 0.76. The confusion matrix indicated strong lesion/background classification, with minor under-segmentation. ICC analysis showed excellent agreement with an experienced surgeon (ICC = 0.85), though lower with a less experienced dentist (ICC = 0.57). The proposed UNet-based model enables efficient segmentation of sequestra, reducing annotation time by >90% while maintaining diagnostic accuracy. The model highlights the clinical utility of AI-assisted decision support in maxillofacial radiology.
Bone sequestration is the formation of a dead bone fragment demarcated from normal bone tissue, usually following dental extractions, delayed healing, chronic infection, or compromised blood supply. It is most frequently associated with osteomyelitis, an infectious process leading to bone inflammation. Osteomyelitis is mostly diagnosed based on clinical presentation with imaging and laboratory evidence but is established through bone biopsy and microbial culture.
Infection of the alveolar bone may be due to various reasons like dental caries, trauma, surgery, and local infection. The dental infections in the majority of instances have localized abscesses, but occasionally, the infection is disseminated, and osteomyelitis follows. Radiation therapy (RT), a standard head and neck cancer therapy, also can lead to osteoradionecrosis (ORN), a disease process in which irradiated bone becomes devitalized, open, and nonhealing. Another mechanism of bone sequestration is medication-related osteonecrosis of the jaw (MRONJ), a complication of antiresorptive and antiangiogenic therapy. First described in 2003 as bisphosphonate-related osteonecrosis of the jaw (BRONJ), the disease was later reclassified by the American Association of Oral and Maxillofacial Surgeons (AAOMS) in 2014 to include cases with non-bisphosphonate therapy, denosumab, and antiangiogenic agents. The medications, given chronically to treat osteoporosis and cancer-related bone disease, lead to interference of vascular supply and thus result in ischemia, hypoperfusion, and eventually bone death.
Although MRONJ is a rare complication, it significantly affects the quality of life of the patient by causing chronic bone exposure, pain, and functional disability. Bone sequestration is a significant problem in the context of all categories of bone disease, including osteomyelitis, ORN, and MRONJ. Pathogenesis, risk factors, and diagnostic criteria should be familiarized with to improve patient outcomes and minimize complications of bone necrosis.
Precise definition of bone sequestrum is essential for diagnosis, treatment planning, and surgical decision-making. Orthopantomography (OPG) or panoramic radiography is widely used in oral and maxillofacial radiology to assess bone disease, such as sequestrum formation. Manual segmentation of sequestrum in OPG is challenging because of anatomical structure superimposition, varying radiopacity, and subjective.
Artificial intelligence (AI) has rapidly expanded in the medical imaging field by providing automated and objective solutions to segmentation challenges. Solutions based on deep learning in the form of CNNs and U-Net-type architectures have been found to be high in accuracy to segment pathological structures from radiographic images. The techniques utilize large data sets to train to identify complex patterns and to differentiate between pathologic areas of bone and normal tissue, potentially improving the capacity to identify sequestrum on OPGs.
Although AI-based segmentation has been comprehensively explored in CT and MRI in most medical centers, it is comparatively less developed in the context of OPG imaging. Since OPG is more accessible and cheaper than cross-sectional imaging modalities, AI-based segmentation of OPGs has the potential to make diagnostic procedures and decision-making in dental and maxillofacial practice more efficient.Variable image quality, annotation challenges, and model generalizability across populations remain barriers that must be addressed before clinical deployment. This study presents an AI-based approach for automated sequestrum segmentation from OPG radiographs, describes the deep learning methodology employed, evaluates model performance, and discusses future clinical applications in dental radiology.
Inclusion Criteria:
Exclusion Criteria: