Cross-Modality AI Fusion of Panoramic Radiographs and CBCT for Patient-Specific Prediction of Hidden Dental Pathology

Abstract

Hidden dental pathologies may remain undetected during routine clinical examinations because of the limitations of conventional dental imaging. Panoramic radiography provides a broad overview of dental and maxillofacial structures, whereas cone-beam computed tomography (CBCT) offers detailed three-dimensional anatomical information. This article explores a proposed cross-modality artificial intelligence (AI) framework for integrating panoramic radiographs and CBCT to support patient-specific prediction of hidden dental pathology. The approach combines deep learning, radiomic feature extraction, anatomical registration, and multimodal feature fusion to identify suspicious lesions and estimate their probability. Potential applications include the detection of occult periapical lesions, root abnormalities, and localized bone defects. Diagnostic performance may be evaluated using sensitivity, specificity, area under the receiver operating characteristic curve, calibration, and external validation. The framework also considers explainability, image quality, radiation safety, and clinical integration. Although advances in dental AI and maxillofacial image analysis provide a foundation for this approach, its diagnostic effectiveness requires systematic experimental evaluation and prospective clinical validation. Cross-modality AI fusion may ultimately support more accurate interpretation, personalized diagnostic assessment, and improved clinical decision-making when used alongside professional judgment and appropriate imaging indications.