Deep Learning for Automated Visualization and Measurement of Dental Plaque Using 3D Intraoral Scanning—A Cross-Sectional Study

Abstract

<p><strong>Background and Aim</strong>: The management of dental plaque is of great importance for the prevention of plaque-associated oral diseases (periodontitis and caries). However, due to their transparent and colorless nature, dental biofilms are challenging to identify visually. Additionally, their irregular shape and distribution make quantification difficult, limiting the efficacy of plaque control and oral health improvement. This study aimed to develop a reliable method for automated visualization and quantification of dental plaque based on deep learning algorithms in 3D intraoral scans.</p><p><strong>Methods</strong>: This was a cross-sectional study on a convenience sample of 110 adult dental hospital attendees. Plaque was disclosed with a two-tone disclosing solution, followed by a full mouth intra oral scanning and a full mouth plaque assessment using the modified O’Leary plaque score index by 6 trained and calibrated dentists. The high inter-examiner reliability was indicated by a Cohen’s kappa value of 0.838. The intra oral scans were annotated by another calibrated examiner who is blind to the clinical examination. A deep learning model was performed to quantify the plaque, and its accuracy was assessed by the intersection over Union (IoU) and the Dice Coefficient (Dice IoU).</p><p><strong>Results</strong>: Overall, the dental plaque coverage on the full dentition ranged from 4.3% to 73.4%, with the mean 29.3% and standard deviation of 10.6%. The IoU for detecting plaque on intra oral scans was 0.584 and the Dice IoU was 0.731.</p><p><strong>Conclusions: </strong>The deep learning model demonstrated a clinically acceptable performance in measuring dental plaque in a 3D structure of the full dentition, reducing the time and human involvement required. This discovery highlights the potential of AI technology in enhancing periodontal health.</p>