ABSTRACT
Background: There is information missing in the literature about the comparison of dentists vs. artificial intelligence (AI) based on diagnostic capability. The aim of this study is to evaluate the diagnostic performance based on radiological diagnoses regarding caries and periapical infection detection by comparing AI software with junior dentists who have 1 or 2 years of experience, based on the valid determinations by specialist dentists. Methods: In the initial stage of the study, 2 specialist dentists evaluated the presence of caries and periapical lesions on 500 digital panoramic radiographs, and the detection time was recorded in seconds. In the second stage, 3 junior dentists and an AI software performed diagnoses on the same panoramic radiographs, and the diagnostic results and durations were recorded in seconds. Results: The AI and the three junior dentists, respectively, detected dental caries at a sensitivity (SEN) of 0.907, 0.889, 0.491, 0.907; a specificity (SPEC) of 0.760, 0.740, 0.454, 0.696; a positive predictive value (PPV) of 0.693, 0.470, 0.155, 0.666; a negative predictive value (NPV) of 0.505, 0.415, 0.275, 0.367 and a F1-score of 0.786, 0.615, 0.236, 0.768. The AI and the three junior dentists respectively detected periapical lesions at an SEN of 0.973, 0.962, 0.758, 0.958; a SPEC of 0.629, 0.421, 0.404, 0.621; a PPV of 0.861, 0.651, 0.312, 0.648; a NPV of 0.689, 0.673, 0.278, 0.546 and an F1-score of 0.914, 0.777, 0.442, 0.773. The AI software gave more accurate results, especially in detecting periapical lesions. On the other hand, in caries detection, the underdiagnosis rate was high for both AI and junior dentists. Conclusions: Regarding the evaluation time needed, AI performed faster, on average.
ABSTRACT
Artificial intelligence (AI) based systems are used in dentistry to make the diagnostic process more accurate and efficient. The objective of this study was to evaluate the performance of a deep learning program for detection and classification of dental structures and treatments on panoramic radiographs of pediatric patients. In total, 4821 anonymized panoramic radiographs of children aged between 5 and 13 years old were analyzed by YOLO V4, a CNN (Convolutional Neural Networks) based object detection model. The ability to make a correct diagnosis was tested samples from pediatric patients examined within the scope of the study. All statistical analyses were performed using SPSS 26.0 (IBM, Chicago, IL, USA). The YOLOV4 model diagnosed the immature teeth, permanent tooth germs and brackets successfully with the high F1 scores like 0.95, 0.90 and 0.76 respectively. Although this model achieved promising results, there were certain limitations for some dental structures and treatments including the filling, root canal treatment, supernumerary tooth. Our architecture achieved reliable results with some specific limitations for detecting dental structures and treatments. Detection of certain dental structures and previous dental treatments on pediatric panoramic x-rays by using a deep learning-based approach may provide early diagnosis of some dental anomalies and help dental practitioners to find more accurate treatment options by saving time and effort.
ABSTRACT
Purpose: The aim of this study was to assess the performance of a deep learning system for permanent tooth germ detection on pediatric panoramic radiographs. Materials and Methods: In total, 4518 anonymized panoramic radiographs of children between 5 and 13 years of age were collected. YOLOv4, a convolutional neural network (CNN)-based object detection model, was used to automatically detect permanent tooth germs. Panoramic images of children processed in LabelImg were trained and tested in the YOLOv4 algorithm. True-positive, false-positive, and false-negative rates were calculated. A confusion matrix was used to evaluate the performance of the model. Results: The YOLOv4 model, which detected permanent tooth germs on pediatric panoramic radiographs, provided an average precision value of 94.16% and an F1 value of 0.90, indicating a high level of significance. The average YOLOv4 inference time was 90 ms. Conclusion: The detection of permanent tooth germs on pediatric panoramic X-rays using a deep learning-based approach may facilitate the early diagnosis of tooth deficiency or supernumerary teeth and help dental practitioners find more accurate treatment options while saving time and effort.
ABSTRACT
This study aims to compare the bioactivity of Biodentine, ProRoot MTA and NeoMTA Plus with regard to their element uptake (Ca, Si and Ca/P) by root canal dentine in a simulated apex (n = 30 each) and evaluate the correlation between the dentine fracture resistance (n = 30 each) and interfacial layer thickness. Specimens immersed in a corrected simulated body solution (c-SBF) for 1, 30 and 90 days were used. In all test materials, the Ca and Si concentrations in the root dentine were found to be significantly higher, whereas the Ca/P and Si concentrations increased over time (P < 0.05). The dentine fracture resistance showed a difference at only day 30. The dentine fracture resistance of Biodentine and ProRoot MTA was positively correlated with the Si and Ca/P values, and the mean interfacial layer thickness of all specimens. A high biomineralisation capacity of ProRoot MTA and Biodentine, and their positive effects on the dentine fracture resistance during the first 30 days suggest that they may present more advantages than NeoMTA Plus in apexification treatment.
