The preliminary in vitro study and application of deep learning algorithm in cone beam computed tomography image...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Radiology

Background:

• Accurate identification of dental implants (bone tissue implants replacing natural tooth roots) is essential for effective repair and maintenance.
• Deep learning excels in image analysis, offering potential for automated identification and classification tasks.

Purpose of the Study:

• To evaluate the efficacy of various deep learning algorithms in recognizing and categorizing three-dimensional (3D) Cone Beam Computed Tomography (CBCT) images of dental implants.
• To determine the best-performing algorithm for an artificial intelligence-based identification system for dental implants.

Main Methods:

• Acquired 3D CBCT imaging data from 27 dental implants of diverse brands and sizes.
• Processed the 3D data into a dataset of 13,500 two-dimensional images.
• Applied nine deep learning algorithms (GoogleNet, InceptionResNetV2, InceptionV3, ResNet50, ResNet50V2, ResNet101, ResNet101V2, ResNet152, ResNet152V2) for image classification and evaluated performance using accuracy, confusion matrix, ROC curve, AUC, model parameters, and training times.

Main Results:

• All nine deep learning algorithms demonstrated satisfactory performance in implant identification.
• ResNet152V2 achieved the highest test accuracy (99.3%), classification accuracy, and area under the curve (AUC) values (1.00).
• The study achieved high training accuracy rates (up to 100%) and test accuracy rates (up to 99.3%) across the evaluated algorithms.

Conclusions:

• Deep learning algorithms, especially ResNet152V2, are highly effective for accurately identifying dental implant brands and specifications from 3D CBCT images.
• An AI identification system utilizing ResNet152V2 offers efficient, accurate, stable, and low-cost recognition of dental implants.
• This technology can significantly aid in the precise repair and maintenance of dental implants.