Deep Learning-Based Analysis of Panoramic and CBCT Images to detect Relations and the Risk of Mandibular third and Inferior Alveolar Nerve: A Scoping Review

Abstract

Abstract Aim: The present scoping review assessed the ability of deep learning (DL) models to determine the relationship between the third mandibular molar (MM3) and the inferior alveolar nerve (IAN) using panoramic radiographs and cone-beam computed tomography (CBCT). The aim was to examine the potential of these models to enhance dental diagnoses of MM3 risk to the nerve. Methods: A systematic literature search encompassed studies published between January 2015 and September 2025 in PubMed, Wiley Online Library, and ScienceDirect. The review included 27 studies which used DL algorithms including the convolutional neural networks, You Only Look Once, ResNet50, AlexNet, and RetinaNet to detect and analyze images. Results: The results showed that these models have high accuracy in predicting anatomical relationships with increased diagnostic efficiency compared with manual methods, reducing nerve complication risks from the MM3. The datasets included between 74 and 4516 X-ray images. The performance metrics included accuracy, F1 score, specificity and sensitivity (Recall). Data heterogeneity including image type, DL architecture, and dataset size makes direct comparison of model performance across studies challenging. Conclusion: DL methods in panoramic radiography and CBCT analysis can improve the accuracy and the efficiency of analyzing the relationship between the MM3 and the IAN, aiding clinical diagnosis and reducing nerve compression risk from wisdom teeth.

Access to Document

Other files and links

• Link to publication in Scopus
• Open Access Version Available

Fingerprint