Accuracy of Computer-Aided Evaluation of the Relationship between Mandibular Third Molar and Mandibular Canal on CBCT Images using Deep Learning Model (Artificial Intelligence): Diagnostic Accuracy Study.

Abstract

Aim: The purpose of this study was to assess the accuracy of a newly developed deep learning model in automatic evaluation of the relationship between mandibular third molar (M3M) and the mandibular canal (MC) on cone beam computed tomography (CBCT) images by comparing it with experienced radiologist opinion.

Methodology: CBCT scans of 184 patients were imported to 3d slicer software. The Radiologist-dependent MC - M3M relation was performed on the axial cuts and then classified into 3 classes: cancellous bone separation, contact with intact cortex and contact with interrupted cortex by 2 Oral and Maxillofacial Radiologists (OMFR) and this classification serves as the ground truth. The annotated data was divided into two groups: 80% for training and validation and 20 % for testing. The data was used to develop the AI model in based on CNN. Confusion matrix and receiver-operating characteristic (ROC) analysis were used in the statistical evaluation of the results of the classification model.

Results: The Average accuracy, precision , recall and F1 score for testing was 0.79, 0.77, 0.77 and 0.77 respectively, while for training was 0.9, 0.91, 0.9 and 0.9 respectively.

Conclusion: Our deep learning model based on CNN showed outstanding performance in the evaluation of the relation between MC and M3M on CBCT images. . However, further development is needed with high quality data to improve the algorithm and validate the accuracy using external validation data sets.