AI-based Prediction of Root Canal Failure Using Smartphone Photographs of Temporary Restorations Combined with Patient-reported Symptoms

Abstract

Root canal treatment is a highly successful procedure when proper diagnosis, disinfection, and coronal sealing are achieved. However, failure may occur because of temporary restoration breakdown, coronal leakage, persistent microbial infection, or delayed recognition of postoperative complications. Conventional follow-up often depends on scheduled clinical visits, which may postpone the detection of early signs of treatment failure. Recent advances in artificial intelligence and smartphone technology offer an opportunity for remote monitoring through automated analysis of clinical photographs and patient-reported symptoms. This approach combines visual assessment of temporary restorations with information such as pain intensity, swelling, biting discomfort, and sensitivity to identify patients at increased risk of treatment failure. Deep learning algorithms can recognize restoration defects, discoloration, fractures, and soft tissue changes from smartphone images, while machine learning models integrate these findings with symptom data to generate individualized risk predictions. Such systems have the potential to improve tele-endodontic care by supporting timely clinical intervention, reducing unnecessary emergency visits, and enhancing patient engagement during the healing period. Although challenges remain regarding image standardization, data privacy, and model validation, AI-assisted remote surveillance represents a promising adjunct to conventional postoperative assessment. Continued refinement of predictive algorithms and mobile health applications may contribute to more efficient follow-up strategies and improved long-term outcomes in endodontic practice.