An Improvised Approach Using YOLOv3 Architecture for Digital Panoramic Teeth Recognition and Classification

Area of Science:

• Computer Vision
• Artificial Intelligence in Dentistry
• Medical Imaging Analysis

Background:

• Access to diagnostic imaging and dental radiographs is crucial for treating oral health issues, yet often limited.
• Tooth loss can result from inadequate diagnostic capabilities.
• Artificial intelligence (AI) offers potential solutions for enhancing dental diagnostic accuracy.

Purpose of the Study:

• To develop an improved AI model for enhanced teeth recognition and classification from dental radiographs.
• To address limitations in existing models for more accurate dental image analysis.
• To explore the potential of AI in early detection of periodontal issues.

Main Methods:

• A modified YOLOv3 model was developed, featuring a smaller convolution layer and architectural deepening for superior feature extraction.
• The enhanced model incorporates a reduction in convolution layers to enable faster recognition.
• Performance was validated using a dedicated test dataset in collaboration with the Radiology department at DAPM RV Dental College and Hospital, Bengaluru.

Main Results:

• The proposed AI model demonstrates improved performance in teeth recognition and classification compared to standard YOLOv3.
• Architectural modifications led to enhanced feature extraction capabilities.
• The study validates the feasibility of using AI for analyzing dental radiographs.

Conclusions:

• The developed AI model shows promise for accurate teeth recognition and classification in dental diagnostics.
• Radiologists play a critical role in training AI systems for reliable clinical application.
• AI systems have the potential to detect periodontal issues from panoramic dental data, aiding in preventive care.