Deep Learning Models for Classification of Deciduous and Permanent Teeth From Digital Panoramic Images

Area of Science:

• Artificial Intelligence in Dentistry
• Digital Radiography Analysis
• Pediatric Dental Diagnostics

Background:

• Dental radiographs are crucial for diagnosis, age determination, and treatment planning.
• Digital radiography and AI offer potential to improve dental diagnostics and reduce clinician workload.
• AI systems can streamline decision-making through classification and localization tasks.

Purpose of the Study:

• To evaluate the performance of various Convolutional Neural Network (CNN) architectures for classifying deciduous and permanent teeth.
• To identify the most effective AI models for automated tooth classification in pediatric dental radiographs.

Main Methods:

• Trained CNN models (Resnet, AlexNet, EfficientNet) on 620 panoramic radiographs from children aged 4-16 years.
• Utilized a 70:15:15 data split for training, validation, and testing.
• Classified teeth as deciduous or permanent using the developed AI models.

Main Results:

• EfficientNetB0 and EfficientNetB3 demonstrated superior performance in tooth classification.
• Both EfficientNet models achieved 98% accuracy, precision, recall, and F1 scores.
• High performance indicates strong capability in identifying dental features for classification.

Conclusions:

• EfficientNetB0 and EfficientNetB3 are highly effective for automated deciduous and permanent tooth classification.
• AI integration shows promise for enhancing diagnostic accuracy in pediatric dentistry.
• These models can potentially improve efficiency and reduce stress for dental professionals.