Transfer learning in a deep convolutional neural network for implant fixture classification: A pilot study.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Dentistry
• Medical Imaging Analysis

Background:

• Accurate classification of dental implant fixtures is crucial for treatment success.
• Deep convolutional neural networks offer potential for automating image analysis tasks in dentistry.
• Transfer learning can enhance model performance with limited datasets.

Purpose of the Study:

• To evaluate the performance of transfer learning using a deep convolutional neural network (YOLOv3) for classifying dental implant fixtures.
• To assess the accuracy of the model in differentiating between various implant systems based on periapical radiographs.

Main Methods:

• A dataset of 355 periapical radiographs featuring Superline, TS III, and Bone Level Implant fixtures was utilized.
• The dataset was split into training (80%) and testing (20%) sets.
• The YOLOv3 deep convolutional neural network was trained using transfer learning for 100, 200, and 300 epochs, with performance evaluated by sensitivity, specificity, and accuracy.

Main Results:

• The YOLOv3 model achieved the highest performance when trained for 200 epochs, with an overall accuracy of 96.7%, sensitivity of 94.4%, and specificity of 97.9%.
• The model demonstrated exceptional performance in classifying Bone Level Implant fixtures, reaching 100.0% for sensitivity, specificity, and accuracy.
• Confidence scores were also highest at 200 epochs, indicating robust classification.

Conclusions:

• Transfer learning enables high-performance classification of implant fixtures using deep convolutional neural networks like YOLOv3, even with limited data.
• The study highlights the potential of AI-driven image analysis for improving the accuracy and efficiency of dental implant assessments.