Abstract

Chronic periodontal disease is a prevalent and progressive condition that significantly affects oral health. Recent advancements in artificial intelligence (AI) offer promising potential for enhancing diagnostic accuracy and classification of periodontal disease. This systematic review aimed to evaluate the current evidence regarding the use of AI in diagnosing and classifying chronic periodontal disease. This review adhered to the PRISMA principles. A thorough literature review was performed till July 7, 2025. Studies were considered if they evaluated AI applications for diagnosing or classifying chronic periodontitis in humans and gave performance measures like accuracy, sensitivity, specificity, or area under the curve (AUC). Data extraction and quality evaluation were conducted. A total of 315 records were identified, with 11 studies meeting the inclusion criteria. These studies used various AI models, including convolutional neural networks (CNN), U-Net architectures, Mask R-CNN, MobileNetV2, and YOLO-based object detectors, applied to data from intraoral photographs, panoramic radiographs, salivary biomarkers, and clinical records. Reported accuracies ranged from 78% to 98%, with AUC values up to 0.96. Most studies demonstrated high consistency between AI and expert human evaluations, especially for radiographic image analysis and staging. However, performance varied by data type, model architecture, and disease severity levels. AI-based systems demonstrated promising diagnostic and classification capabilities for chronic periodontal disease, with performance metrics comparable to expert clinicians. Deep learning models, particularly those applied to radiographic imaging, showed strong potential for integration into routine dental practice. However, further validation on diverse datasets and real-world clinical settings is needed before widespread implementation.

Key words: Artificial intelligence, chronic periodontitis, deep learning, machine learning, diagnosis, systematic review.