AI-Based Detection of Dental Features on CBCT: Dual-Layer Reliability Analysis

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) holds potential for improving diagnostic accuracy in cone-beam computed tomography (CBCT) analysis.
• Current validations often focus on single-tooth assessments, lacking evaluation of full-mouth diagnostic outcomes.

Purpose of the Study:

• To assess the diagnostic accuracy of a commercial AI platform for identifying dental treatment features on CBCT images.
• Evaluation was conducted at both the individual tooth and full-mouth scan levels.

Main Methods:

• A retrospective analysis of 147 CBCT scans (4704 tooth positions) was performed.
• Two expert readers established consensus annotations as the reference standard.
• An AI system processed anonymized scans, with diagnostic metrics calculated using patient-clustered bootstrap confidence intervals.

Main Results:

• The AI platform achieved excellent tooth-level accuracy (>99%) for most dental features.
• Highest sensitivity was observed for missing teeth (99.3%) and endodontic treatments (99.0%).
• Full-scan level "Perfect Agreement" was achieved in 82.3% of cases, with errors often related to teeth with multiple findings.

Conclusions:

• The AI platform demonstrates high accuracy for detecting isolated dental features on CBCT scans.
• Its reliability is moderate for generating complete full-mouth reports.
• The AI functions effectively as a diagnostic aid rather than an autonomous system.