تشخیص پوسیدگی های عمیق اینترپروگزیمال و بررسی ارتباط آن با پالپ در رادیوگرافی دیجیتال بایت وینگ با استفاده از مدل های هوش مصنوعی
کد: G-1428
نویسندگان: Mohammad Hadi Goldani * ℗, Atefeh Gohari, Abdalsamad Keramatfar, Farnoosh Alimohammadi
زمان بندی: زمان بندی نشده!
برچسب: سیستم های تصمیم یار بالینی
دانلود: دانلود پوستر
خلاصه مقاله:
خلاصه مقاله
Background and aims: Intraoral radiographs are essential for assessing interproximal caries, yet their diagnostic reliability is often compromised by intra- and inter-examiner variability. Artificial intelligence (AI), particularly convolutional neural networks (CNNs), presents a viable solution to reduce human diagnostic errors by enhancing accuracy. This study aimed to develop and evaluate a CNN-based model as an adjunctive tool for diagnosing deep interproximal caries and assessing their proximity to the dental pulp. Method: A total of 899 bitewing radiographs were selected based on predefined inclusion and exclusion criteria. An endodontist classified the radiographs into five color-coded categories according to caries severity: enamel lesions (blue), dentinal lesions extending less than half of the dentin thickness (yellow), deep dentinal lesions extending beyond half the dentin thickness (orange), and lesions reaching the pulp (red). The You Only Look Once v8 (YOLOv8) model was implemented using Python (v3) and the Ultralytics framework, with training and testing performed to detect and classify carious lesions. Results: The model exhibited optimal performance in detecting pulp-involving caries (red group), achieving 85% sensitivity, followed by yellow (71%), blue (61%), and orange (64%) groups. For deep caries extending beyond half the dentin thickness, the model correctly identified lesions with a high likelihood of pulp exposure in 90% of cases and those with low/no likelihood in 75% of cases. Conclusion: The study successfully developed a CNN-based model for detecting and classifying interproximal caries in bitewing radiographs. While the results are promising, detection inaccuracies indicate the need for further refinement in future research to enhance diagnostic precision.
کلمات کلیدی
Bitewing Radiography, Interproximal Caries, Deep Learning
