DeepPlaq: Dental plaque indexing based on deep neural networks

Area of Science:

• Dental diagnostics
• Artificial intelligence in healthcare
• Computer vision applications

Background:

• Dental plaque assessment is crucial for treatment selection.
• Current methods for plaque evaluation can be subjective.
• Objective and accurate plaque indexing is needed.

Purpose of the Study:

• To validate the efficacy of Convolutional Neural Network (CNN) models in assessing dental plaque indices.
• To develop and evaluate an AI-driven system for objective dental plaque evaluation.

Main Methods:

• A three-stage AI approach was used, involving You Look Only Once version 8 (YOLOv8) for tooth detection and Segment Anything Model (SAM) for segmentation.
• A novel dataset of 1400 single-tooth images was created.
• The DeepPlaq multi-class classification model was trained and validated using the Quigley-Hein Index (QHI) scoring system.

Main Results:

• The teeth detection model achieved a mean average precision (mAP) of approximately 0.941 ± 0.005.
• The DeepPlaq model reached a maximum accuracy of 0.84 in plaque indexing, with an average scoring error below 0.25 on a 0-5 scale.
• The AI system demonstrated high accuracy in identifying teeth and segmenting plaque.

Conclusions:

• The developed three-stage AI approach effectively detects and segments teeth for plaque analysis.
• The DeepPlaq model shows significant potential for accurate dental plaque index assessment.
• AI application in dental plaque evaluation can enhance diagnostic precision and treatment efficiency.