Creative Commons source reading · Data quality and algorithm evaluation
An Attention-Enhanced RegNetY Framework for Detection and Classification of Vertical Misfit in Implant-Supported Restorations: A Retrospective Study
Tuba Talo Yildirim; Aybike Cengiz Dagtekin; Nurullah Düger; Ayşe Rençber Kizilkaya; Furkan Talo; Emre Arslan; Mucahit Karaduman; Muhammed Yildirim
Originally published in Diagnostics (Basel) · 2026-05-25
From the article
Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.
Abstract
Background/Objectives: The aim of this study is to test different convolutional neural network (CNN) and Transformer-based models to detect and classify vertical misfit at the abutment-prosthesis interface on panoramic radiographs, and to develop a hybrid deep learning model enhanced with attention mechanisms. Methods: A dataset consisting of a total of 566 images, manually classified as 249 ‘fit’ and 317 ‘misfit’ cases by two experts, was created. Images were resized to 224 × 224 and divided into training, validation, and test groups. The deep learning model yielding the most successful results was determined as the backbone; a hybrid model was developed by integrating three different attention modules (SE, CBAM, and ECA) into this structure. Model performance was evaluated using accuracy, precision, sensitivity, and F1 score…
3. Results
A hybrid deep learning model enhanced with attentional mechanisms has been developed to detect and classify vertical discrepancies at the abutment-prosthesis interface in panoramic radiographs. The developed model was compared with different CNN and ViT models accepted in the literature. The performance of the models was evaluated using different performance measurement metrics. The application results were obtained in the Google Colab environment using an NVIDIA Tesla T4 GPU. The performance metrics of all models used in our study are presented in Table 1. In this study, 4 CNN-based and 2 Transformer-based models were utilized. CNN-based models demonstrated superior performance across all metrics compared to Transformer-based models (DeiT, Swin-Tiny). The models exhibiting the highest accuracy rates, in descending order, were…
4. Discussion
Vertical misfit at the prosthesis-abutment interface leads to complications such as peri-implant mucositis, peri-implantitis, crown loss, and implant loss [4,7]. The aim of this study was to develop a high-precision artificial intelligence model for the detection of vertical misfit at the abutment-prosthesis interface using challenging imaging modalities containing distortion and low resolution, such as panoramic radiographs. The findings indicated that the RegNetY-800MF model reinforced with attention mechanisms (SE, CBAM, ECA) performed better than both traditional CNN and Transformer architectures, achieving an accuracy rate of 99.12%. There are limited studies on the detection of abutment-prosthesis misfit using deep learning. The only current study on this subject, conducted by Fasih et al., achieved 92.7% accuracy using the…
Editorial reading guide
- What data created and tested the model, and was evaluation independent?
- What is the cost of a false positive, false negative or out-of-distribution decision?
