Deep Learning for Periodontitis Diagnosis on Two Dimensional Dental Radiograph: A Systematic Review

Abstrak

Periodontitis is an inflammatory disease that affects the supporting structures of the teeth and is a major contributor to tooth loss. Traditional diagnosis through clinical examination and manual interpretation of two-dimensional (2D) dental radiographs is prone to variability and subjectivity. The emergence of deep learning (DL) offers a powerful tool in medical image analysis, including dental radiography. This study aims to systematically review the existing literature on the use of DL approaches for diagnosing periodontitis using two-dimensional (2D) dental radiographic images, and to assess their diagnostic effectiveness in comparison to conventional clinician-based evaluation. A systematic literature review (SLR) was conducted following the PRISMA 2020 protocol and guided by the PICO framework. Five major databases (Scopus, PubMed, Semantic Scholar, Web of Science, and ScienceDirect) were searched for relevant studies published between 2016 and 2025. A total of 27 studies (across 29 reports) were included based on eligibility criteria, covering classification, segmentation, or detection tasks using panoramic, periapical, or bitewing radiographs. The results indicate that DL models show high diagnostic potential, with classification accuracies often exceeding 80% and segmentation models achieving Dice coefficients above 0.88. Although some models outperformed clinicians, external validation and real-world deployment remain limited. This review highlights both the diagnostic potentials and present limitations of DL in 2D dental radiographs. In conclusion, DL shows substantial promise for automated periodontitis diagnosis using 2D radiographs, though challenges still remain in standardization, external validation, and integration into clinical workflows.