Related Experiment Video
Updated: May 14, 2025
05:49
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
Published on: February 23, 2024
743
Optimized classification of dental implants using convolutional neural networks and pre-trained models with
Reza Ahmadi Lashaki1, Zahra Raeisi2, Nasim Razavi3
1Department of Computer Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran. lashaki@tabrizu.ac.ir.
BMC Oral Health
|April 11, 2025
Summary
Data augmentation and preprocessing significantly improve dental implant classification accuracy. Convolutional Random Forest (CRF) and Convolutional Neural Networks (CNN) were top classifiers, while VGG16 excelled among pre-trained models.
Area of Science:
- Medical Imaging
- Computer Vision
- Dental Technology
Background:
- Accurate dental implant state classification is crucial for diagnosis and treatment planning.
- Radiography images are commonly used but require robust analysis methods.
- Existing classification methods may benefit from advanced preprocessing and augmentation.
Purpose of the Study:
- To evaluate the performance of various classifiers and pre-trained models for dental implant state classification.
- To assess the impact of data augmentation and preprocessing on classification accuracy.
- To identify optimal models for classifying dental implant states from radiography images.
Main Methods:
- A dataset of 511 periapical images was augmented to 5110 using rotation, flipping, and scaling.
- Images underwent preprocessing: resizing, sharpening, noise reduction, contrast enhancement, and implant-specific masking.
- Classifiers included Convolutional Neural Networks (CNN), Convolutional Support Vector Machine (CSVM), Convolutional Decision Tree (CDT), and Convolutional Random Forest (CRF).
- Pre-trained models (VGG16, ResNet50, Xception) were utilized for feature extraction.
- Performance was evaluated using accuracy, precision, recall, F1 score, and ROC AUC with fivefold cross-validation.
Main Results:
- Convolutional Random Forest (CRF) showed high performance for ITI with Bego implants (accuracy 0.8966).
- Convolutional Neural Networks (CNN) achieved the best results for Bicon with Bego implants (accuracy 0.9533).
- VGG16, a pre-trained model, demonstrated superior performance for Bicon vs. ITI classification (accuracy 0.9865) when using preprocessed data.
- Data augmentation and preprocessing significantly enhanced the performance of all evaluated classifiers.
Conclusions:
- Data augmentation and preprocessing are vital for improving the robustness and accuracy of dental implant classification.
- CRF and CNN emerged as highly effective classifiers for this task.
- VGG16 proved to be the most effective pre-trained model, underscoring the value of transfer learning in this domain.
- The study confirms the importance of advanced image processing techniques in dental radiography analysis.
