Artificial Intelligence in Periodontology: A Systematic Review

Area of Science:

• Periodontology
• Artificial Intelligence
• Machine Learning
• Radiology

Background:

• Artificial intelligence (AI) is increasingly explored for applications in periodontology.
• This review focuses on deep learning for image-based diagnosis and other AI applications in periodontal care.

Purpose of the Study:

• To comprehensively review AI applications in periodontology.
• To assess deep learning for image-based periodontitis diagnosis.
• To explore non-image-based AI applications in periodontal care.

Main Methods:

• Systematic review adhering to PRISMA guidelines.
• Searched six major scientific databases.
• Two parts: PICO 1 (deep learning on imaging data) with QUADAS-2 RoB assessment, and PICO 2 (non-imaging data) with narrative synthesis.

Main Results:

• PICO 1: 29 studies, deep learning achieved high accuracy (81%-99%) for periodontitis classification on radiographs.
• PICO 2: 65 studies, AI demonstrated effectiveness in diagnosis, risk stratification, progression modeling, and clinical data mining.
• AI applications showed promising metrics, but risk of bias and applicability concerns were noted, particularly regarding dataset diversity.

Conclusions:

• Generalizability is a key limitation due to data diversity, inconsistent metrics, and lack of external testing.
• Future research should focus on multicenter evaluations and transparent reporting.
• Prospective studies assessing workflow impact and patient outcomes are needed.