Efficacy of deep learning models and dental professionals in identifying dental implants

深度学习模型和牙科专业人员在识别牙科植入物方面的有效性

阅读:1

Abstract

PURPOSE: Implant identification is a pressing concern in dental implantology, and artificial intelligence (AI) has been evaluated for this purpose. YOLO, a state-of-the-art object detection model, is suitable for medical imaging; therefore, this study assessed YOLOv11-the latest iteration-for identifying 10 implant types in Indian clinical settings and compared its accuracy to that of dental professionals. MATERIALS AND METHODS: A dataset of 3,161 radiographs, comprising both periapical and panoramic images of 10 implant types, was annotated and used to train and test YOLOv11. Training was performed on Google Colab using an NVIDIA Tesla T4 GPU (16 GB VRAM). A random sample of 200 radiographs was selected from the test dataset and presented to 50 dental practitioners for implant identification. Their responses were analysed and compared, using the chi-square test for statistical significance. RESULTS: YOLOv11 achieved precision of 0.87, recall of 0.85, an F1-score of 0.86, and an mAP50 of 0.899. The model achieved excellent classification accuracy for Adin (95%), MIS (94%), Bego (92%), ITI (96%), and Bicon (97%). Moderate accuracy was noted for Noris (82%), Osstem (85%), AlphaBio (88%), Dentium (77%), and Bioline (75%). YOLOv11 demonstrated higher overall accuracy and consistency than dental professionals. Dentists' accuracy ranged from 27% to 49%, whereas that of YOLOv11 ranged from 92% to 100%. CONCLUSION: YOLOv11 recognised most implant classes with over 90% accuracy, surpassing traditional manual techniques in implant detection. Although the model is dependable and efficient, certain aspects require improvement. The study also emphasises the significance of a region-specific approach for clinical relevance.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。