RESUMEN
STATEMENT OF PROBLEM: Dental implant systems can be identified using image classification deep learning. However, investigations on the accuracy of classifying and identifying implant design through an object detection model are lacking. PURPOSE: The purpose of this study was to evaluate the performance of an object detection deep learning model for classifying the implant designs of 103 types of implants. MATERIAL AND METHODS: From panoramic radiographs, 14 037 implant images were extracted. Implant designs were subdivided into 10 classes in the coronal, 13 in the middle, and 10 in the apical third. Classes with fewer than 50 images were excluded from the training dataset. Among the images, 80% were used as training data, and the remaining 20% as test data; the data were generated 3 times for 3-fold cross-validation (implant datasets 1, 2, and 3). Versions 5 and 7 of you only look once (YOLO) algorithm were used to train the model, and the mean average precision (mAP) was evaluated. Subsequently, data augmentation was performed using image processing and a real-enhanced super-resolution generative adversarial network, and the accuracy was re-evaluated using YOLOv7. RESULTS: The mAP of YOLOv7 in the 3 datasets was 0.931, 0.984, and 0.884, respectively, which were higher than the mAP of YOLOv5. After image processing in implant dataset-1, the mAP improved to 0.986 and, with the real-enhanced super-resolution generative adversarial network, to 0.988 and 0.986 at magnification ×2 and ×4, respectively. CONCLUSIONS: The object detection model for classifying implant designs found a high accuracy for 26 classes. The mAP of the model differed depending on the type of algorithm, image processing process, and detailed implant design.
RESUMEN
Purpose: To evaluate the accuracy and clinical usability of an identification model using ensemble deep learning for 130 dental implant types. Materials and Methods: A total of 28,112 panoramic radiographs were obtained from 30 domestic and foreign dental clinics. From these panoramic radiographs, 45,909 implant fixture images were extracted and labeled based on electronic medical records. Dental implants were classified into 130 types according to the manufacturer, the manufacturer's implant system, and the diameter and length of the implant fixture. Regions of interest were manually cropped, and data augmentation was performed. According to the minimum number of images collected per implant type, the datasets were classified into three sets: an overall total of 130 and two subsets that consisted of 79 and 58 types. EfficientNet and Res2Next algorithms were used for image classification in deep learning. After testing the performance of the two models, the ensemble learning technique was applied to improve accuracy. The top-1 accuracy, top-5 accuracy, precision, recall, and F1 scores were calculated according to algorithms and datasets. Results: For the 130 types, the top-1 accuracy, top-5 accuracy, precision, recall, and F1 scores were 75.27, 95.02, 78.84, 75.27, and 74.89, respectively. In all cases, the ensemble model performed better than EfficientNet and Res2Next. When using the ensemble model, the accuracy increased as the number of types decreased. Conclusion: The ensemble deep learning model for the identification of 130 types of dental implants showed higher accuracy than the existing algorithms. To further improve the performance and clinical usability of the model, images with higher quality and fine-tuned algorithms optimized for implant identification are required.