The Impact Of Artificial Intelligence (AI) In Various Dental Specialties: A Systematic Review

Abstract

Background:
Artificial intelligence (AI), including machine learning and deep learning approaches, is increasingly integrated into dental diagnostics and treatment planning. Across dental specialties, AI has shown potential to enhance diagnostic accuracy, optimize workflows, and support clinical decision-making. However, the breadth of applications and the performance consistency of AI systems remain variable, necessitating a comprehensive synthesis of current evidence.

Objective:
To systematically evaluate the impact, diagnostic performance, and clinical utility of artificial intelligence applications across various dental specialties.

Methods:
A systematic review was conducted following PRISMA guidelines. A comprehensive search of PubMed, ScienceDirect, Google Scholar, and the Cochrane Library was performed to identify studies assessing AI applications in dentistry. Eligible studies included primary research evaluating AI-based diagnostic, predictive, or treatment-planning tools across dental specialties. Data extraction encompassed study design, AI model characteristics, dataset size, performance metrics, and key findings. Quality assessment was performed using QUADAS-2 for diagnostic studies, the Newcastle–Ottawa Scale for observational studies, and the Cochrane Risk of Bias tool for trials. Due to significant heterogeneity in methodologies, datasets, and performance indices, a narrative synthesis was conducted instead of a meta-analysis.

Results:
Fifteen studies met the inclusion criteria. AI applications were most commonly studied in periodontology (n=5) and orthodontics (n=5), followed by endodontics (n=3), cariology/radiology (n=2), and paediatric dentistry (n=1). Deep learning architectures—particularly CNNs, U-Net models, Faster R-CNN, Mask R-CNN, and YOLO variants—demonstrated strong diagnostic performance in tasks such as detecting periodontal bone loss (AUC up to 0.951), segmenting caries lesions (accuracy up to 0.93), identifying cephalometric landmarks (2 mm success rates >90%), and detecting periapical pathology or fractured instruments. Several AI systems matched or exceeded clinician performance, especially in radiographic interpretation. However, limitations included small datasets, lack of external validation, variability in ground truth labeling, and reduced performance for subtle or complex pathologies.

Conclusion:
AI demonstrates significant promise across multiple dental specialties, showing high diagnostic accuracy and strong potential as a clinical decision-support tool. Despite encouraging results, widespread adoption is limited by methodological heterogeneity, insufficient validation, and inconsistent reporting standards. Future research should emphasize larger, multi-center datasets, external validation, transparency in model development, and real-world clinical trials to strengthen evidence for clinical integration.