Abstract
BACKGROUND: Periodontal disease is among the most prevalent diseases globally, yet it often goes undetected, with only 27% of cases reported to receive treatment in the US. In situations when clinical examination is not possible, radiographic findings may serve as an alternative method for periodontal disease diagnosis. Several Artificial Intelligence (AI) platforms were developed to detect the amount of radiographic bone loss with different accuracy levels. Therefore, this study aimed to validate the Overjet AI platform for diagnosing periodontal disease using full-mouth radiographs against the gold standard of clinical-radiographic analysis, and to compare its diagnostic accuracy with manual radiographic assessment by a general practitioner (GP). METHODS: In this study radiographic records of patients aged over 29 years were utilized to validate the use of radiographic analysis (using full-mouth radiographs) solely by GP and by Overjet AI software for detecting periodontal disease. A sample size calculation was performed with 95% power and an alpha of 0.05 to identify an effect size of 0.2. To evaluate the diagnostic accuracy of the Overjet AI software, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated against the gold standard of the combined clinical and radiographic analysis conducted by a periodontist. Cohen’s Kappa agreement was used to compare results from manual radiographic analysis by GP verses AI platform. RESULTS: The study included radiographic records of 103 patients. Results showed that detecting periodontal disease across different severity level using radiographic analysis manually achieved 100 − 83% sensitivity and 94 − 90% specificity. While Overjet AI software achieved 100 − 82% sensitivity and 96 − 89% specificity. The Cohen’s Kappa agreement between the GP and AI platform results was between 0.49 and 0.85 representing a moderate to almost perfect agreement. CONCLUSION: AI-based radiographic analysis offers a rapid and accurate alternative to manual dentist assessments, particularly for detecting moderate to severe periodontal disease. The findings suggest a potential for integrating AI technology with conventional clinical exam to improve the efficiency and accuracy of periodontal disease detection and monitoring.