AI-Assisted Detection of Secondary Caries and Assessment of Composite Restoration Failure: Association with Periodontal and Oral Biological Factors

Abstract

Background: Secondary caries and degradation of resin based composite restorations continue to be significant problems for restoration repair and replacement. Besides the material consideration, there are a variety of factors that can affect long-term performance of the composite restorations including periodontal inflammation, plaque accumulation, salivary characteristics, and cariogenic microorganisms. AI supported diagnostic systems could be available to assist in the detection of recurrent caries and the assessment of risk of failure of restorations.

Objective: To evaluate AI-assisted detection of secondary caries and assess composite restoration failure in relation to periodontal, oral biological, and restoration-related factors.

Methods: This analytical cross-sectional study included 85 participants, with one eligible composite restoration evaluated per participant, resulting in 85 restorations for analysis. Clinical, periodontal, radiographic, salivary, and microbiological parameters were recorded. Secondary caries was assessed using combined clinical and radiographic examination, while AI-assisted radiographic assessment was performed using Diagnocat™ for detection of secondary caries. Composite restoration failure was determined clinically using predefined criteria. Plaque Index, Gingival Index, bleeding on probing, probing pocket depth, salivary pH, salivary flow rate, buffering capacity, Streptococcus mutans level, restoration age, number of restored surfaces, and marginal adaptation were evaluated. Diagnostic performance measures and multivariable logistic regression were used for statistical analysis.

Results: Secondary caries was confirmed in 29 of 85 restorations (34.1%), while 25 restorations (29.4%) were clinically classified as failed. AI-assisted secondary caries detection demonstrated a sensitivity of 89.7%, specificity of 91.1%, positive predictive value of 83.9%, negative predictive value of 94.4%, and overall accuracy of 90.6%. Restorations with secondary caries were associated with significantly higher Plaque Index, Gingival Index, and probing pocket depth, along with lower salivary pH and flow rate. High S. mutans levels were also significantly more frequent in restorations affected by secondary caries. Composite restoration failure was significantly associated with secondary caries, restoration age >5 years, involvement of three or more surfaces, poor marginal adaptation, poor oral hygiene, and high S. mutans levels. In multivariable analysis, secondary caries, poor marginal adaptation, restoration age >5 years, and high S. mutans levels remained independently associated with composite restoration failure.

Conclusion: AI-assisted radiographic assessment demonstrated high diagnostic performance for secondary caries detection. Composite restoration failure was independently associated with secondary caries, defective marginal adaptation, greater restoration age, and elevated S. mutans levels. These findings support the use of AI as an adjunct to conventional secondary caries detection while emphasizing the importance of periodontal, biological, and restoration-related factors in the clinical assessment of composite restoration outcomes.