Artificial Intelligence Model to Detect Real Contact Relationship between Mandibular Third Molars and Inferior Alveolar Nerve Based on Panoramic Radiographs

Abstract

This study aimed to develop a novel detection model for automatically assessing the real contact relationship between mandibular third molars (MM3s) and the inferior alveolar nerve (IAN) based on panoramic radiographs processed with deep learning networks, minimizing pseudo-contact interference and reducing the frequency of cone beam computed tomography (CBCT) use. A deep-learning network approach based on YOLOv4, named as MM3-IANnet, was applied to oral panoramic radiographs for the first time. The relationship between MM3s and the IAN in CBCT was considered the real contact relationship. Accuracy metrics were calculated to evaluate and compare the performance of the MM3-IANnet, dentists and a cooperative approach with dentists and the MM3-IANnet. Our results showed that in comparison with detection by dentists (AP = 76.45%) or the MM3-IANnet (AP = 83.02%), the cooperative dentist-MM3-IANnet approach yielded the highest average precision (AP = 88.06%). In conclusion, the MM3-IANnet detection model is an encouraging artificial intelligence approach that might assist dentists in detecting the real contact relationship between MM3s and IANs based on panoramic radiographs.

Keywords: YOLOv4; contact relationship; deep learning network; inferior alveolar nerve; mandibular third molar; panoramic radiograph.

Figure 1

Panoramic view of a patient with corresponding CBCT results. Tooth position was recorded using the Federation Dentaire International system. Forty-eight showed that the dental roots were in contact with the IAN in the panoramic radiograph but not in contact in CBCT, so 48 was classified into the non-contact group. Thirty-eight showed the dental roots in contact with the IAN in both the panoramic radiograph and the CBCT result, so 38 was classified into the contact group.

Figure 2

Model of MM3–IANnet system architecture. In step 3, process 1–process 3 was the upper sampling operation and process 4–process 6 was the lower sampling operation.

Figure 3

Output results of MM3–IANnet, dentists and cooperative dentist–MM3–IANnet approach. Three typical examples are presented. (A) According to the contact relationship in CBCT images (gold standard), the MM3 and IAN were divided into the contact group. The test result of MM3–IANnet was the contact. Two of the five dentists considered the case a contact, and the other three did not. The test result of dentist–MM3–IANnet (voting experiment) was a contact. (B) The MM3 and IAN were divided into the non-contact group. The test result of MM3–IANnet was the non-contact, and two of the five dentists considered the case a contact, while the other three did not. The test result of dentists–MM3–IANnet was the non-contact. (C) The MM3 and IAN were divided into the non-contact group. The test result of MM3-IANnet was a contact, five dentists considered the case non-contact, and the test result of dentist–MM3–IANnet was non-contact.