aceAutomated classification of dental implant brands and prosthetic platform sizes on panoramic radiographs using deep learning
Statement of problem Incomplete clinical records can be a significant hurdle in implant dentistry, transforming routine maintenance or a restorative task into a complex search. When the primary documentation-such as the implant passport or surgical report-is missing, the clinician is forced to rely on radiographic identification and trial-and-error, which increases the risk of component mismatch and patient dissatisfaction. Purpose The purpose of this study was to develop and validate an artificial intelligence…
"NMCP Installs New, Safer Dental Equipment" by NavyMedicine is marked with Public Domain Mark 1.0. To view the terms, visit https://creativecommons.org/publicdomain/mark/1.0/.
Researchers developed and validated a two-stage convolutional neural network pipeline to detect dental implants on panoramic radiographs and classify them by brand and prosthetic platform size. Using 387 radiographs with 1004 implants, the Faster R-CNN with EfficientNet-B7 backbone achieved 99.45% detection accuracy and 85.60% combined brand-and-platform accuracy across 25 partitions.
The work addresses incomplete clinical records that force clinicians to rely on radiographic identification and trial-and-error, increasing mismatch risk. By automating identification, the system could improve diagnostic efficiency and standardization, though the reported results are from a retrospective dataset of under 400 radiographs and have not yet been tested in prospective clinical workflows.
- Dataset included 387 panoramic radiographs with 1004 dental implant images split 80/10/10 across 25 independent partitions.
- Best model was Faster region-based CNN combined with EfficientNet-B7 backbone trained on manually annotated images with ground truth from patient records.
- Brand classification accuracy reached 98.93% for Callus Pro, 97.81% for Denti Root Form, and 98.85% for NobelReplace Conical Connection PMC.
A two-stage CNN pipeline using Faster R-CNN with EfficientNet-B7 detected dental implants on panoramic radiographs and classified brand and prosthetic platform size with high accuracy, supporting faster and more standardized clinical workflows when records are missing.
The rundown
The study used 387 panoramic radiographs containing 1004 implant images, divided with an 80/10/10 stratified split across 25 partitions. Ground truth for brand and platform size came from patient records, and models were trained on manually annotated images.
Detection performance averaged 0.60 false negatives per split. Platform-size accuracy was 87.62% for narrow, 87.78% for regular, and 96.79% for unknown sizes, while overall brand-classification accuracy was 95.66%.
Sources
- Peer-reviewedThe Journal of Prosthetic Dentistry2026-07-28
How should this claim be treated?
The debate