The aim of this study was to investigate if the metal artefact reduction (MAR) tool used in the software of the ORTHOPANTOMOGRAPH(®) OP300 (Instrumentarium Dental, Tuusula, Finland) can improve the gray value levels in post-operative implant scans. 20 potential implant sites were selected from 5 edentulous human dry mandibles. Each mandible was scanned by a CBCT scanner, and images were produced under three different conditions: implant sites drilled but no implants inserted, implants inserted without application of MAR and implants inserted with application of MAR. Using Geomagic(®) Studio 2012 (Geomagic, Morrisville, NC) and 3Diagnosys(®) v. 5.3.1 (3Diemme(®) SRL, Cantù, Italy) software, three scans of each mandible were superimposed. The mean gray value of identical regions of bone around the implants was derived for each condition. The differences between gray value measurements at implant sites derived from different conditions were assessed. A significant difference was found between mean gray values from the scans with no implants inserted and with implants inserted (with and without MAR) (p = 0.012). No significant difference was revealed for gray values measured from scans with and without MAR (p = 0.975). The MAR tool in the software of the ORTHOPANTOMOGRAPH OP300 CBCT scanner does not significantly correct the voxel gray values affected by the metal artefact in the vicinity of an implant in human dry mandibles.

Background: Artificial intelligence has been proven to improve the identification of various maxillofacial lesions. The aim of the current study is two-fold: to assess the performance of four deep learning models (DLM) in external root resorption (ERR) identification and to assess the effect of combining feature selection technique (FST) with DLM on their ability in ERR identification. Methods: External root resorption was simulated on 88 extracted premolar teeth using tungsten bur in different depths (0.5 mm, 1 mm, and 2 mm). All teeth were scanned using a Cone beam CT (Carestream Dental, Atlanta, GA). Afterward, a training (70%), validation (10%), and test (20%) dataset were established. The performance of four DLMs including Random Forest (RF) + Visual Geometry Group 16 (VGG), RF + EfficienNetB4 (EFNET), Support Vector Machine (SVM) + VGG, and SVM + EFNET) and four hybrid models (DLM + FST: (i) FS + RF + VGG, (ii) FS + RF + EFNET, (iii) FS + SVM + VGG and (iv) FS + SVM + EFNET) was compared. Five performance parameters were assessed: classification accuracy, F1-score, precision, specificity, and error rate. FST algorithms (Boruta and Recursive Feature Selection) were combined with the DLMs to assess their performance. Results: RF + VGG exhibited the highest performance in identifying ERR, followed by the other tested models. Similarly, FST combined with RF + VGG outperformed other models with classification accuracy, F1-score, precision, and specificity of 81.9%, weighted accuracy of 83%, and area under the curve (AUC) of 96%. Kruskal Wallis test revealed a significant difference (p = 0.008) in the prediction accuracy among the eight DLMs. Conclusion: In general, all DLMs have similar performance on ERR identification. However, the performance can be improved by combining FST with DLMs.