What type of study is this?

This is a Experimental Study study.

September 12, 2025Open Access

Development and Performance of an Artificial Intelligence-Based Deep Learning Model Designed for Evaluating Dental Ergonomics

Key Points

The AI model excelled in assessing dental ergonomics, significantly outperforming human evaluators in sensitivity.
Model showed excellent agreement with calibrated evaluators, achieving a Kappa score of 0.922, indicating high reliability.
AI demonstrated sensitivity of 97% compared to just 20.5% for human evaluations, highlighting its effectiveness.
The model's AUC was 0.917, suggesting it may serve as a 'gold standard' for assessing dental operator ergonomics.

Abstract

Background/Objectives: Dental professionals engage in a variety of dental procedures within a confined workspace that is often challenging to access and navigate. This environment frequently results in static, asymmetrical, and inappropriate postures, which can lead to muscular imbalances and cause pain or damage to the musculoskeletal system. Such issues can adversely affect the dental workforce, resulting in increased absenteeism, reduced productivity, disability, and premature retirement from the profession. Therefore, the objective of this study was to develop and evaluate the performance of an Artificial Intelligence (AI)-based deep learning model designed to assess dental ergonomics. Methods: An AI-based Dental Ergonomic Posture Assessment Model SBK-DentErgo was developed through the strategic integration of YOLOv11 and MediaPipe. Model training and validation were conducted using 500 photographs of dental professionals performing procedures on patients, captured from both frontal and sagittal planes. In the initial phase of the study, two calibrated evaluators assessed 50 photographs, demonstrating excellent agreement. In the subsequent phase, five dental specialists, along with the AI model, evaluated the same set of photographs, and the results were recorded. Results: AI-based model demonstrated excellent agreement with that of calibrated evaluators (Kappa = 0.922, p = 0.000). The reliability of AI-based scores was also consistent (ICC = 1.000, p = 0.000). Human evaluation of ergonomic posture exhibited very low sensitivity (20.5%) compared to AI, which showed very high sensitivity (97%). The specificity of human evaluation was also extremely low (9.1%) in contrast to AI (85.7%). The AI model (AUC = 0.917, 95% CI 0.762–1.000) could serve as the ‘gold standard’ in evaluating dental operator ergonomics. Conclusions: This AI model demonstrated exceptional performance in evaluating the working postures of dental professionals, surpassing experienced specialists in both sensitivity and specificity. The model provides real-time feedback, enabling dentists to conduct self-assessments and correct their posture immediately, thereby preventing postural issues.

Ask AI

Helpful

Bookmark

View Full Paper