Accuracy of Artificial Intelligence in Evaluation of the Relationship Between Mandibular Third Molar and Mandibular Canal on CBCT

Artificial Intelligence Clinical Trial

Official title:

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.

Clinical Trial Details — Status: Recruiting

Administrative data

• NCT number: NCT05350228
• Other study ID #: ORAD AI 1-1
• Status: Recruiting
• Start date: May 2022
• Est. completion date: December 2023

Study information

• Verified date: April 2022
• Source: Cairo University
• Contact: Ahmed Salama, Msc
• Phone: +201019932383
• Email: ahmed_magdy[@]dentistry.cu.edu.eg
• Is FDA regulated: No
• Study type: Observational

Clinical Trial Summary

Convolutional neural network (CNN) are computer applications that assist in the detection and/or diagnosis of diseases by providing an unbiased "second opinion" to the image interpreter10, aiming at improving accuracy and reducing time for analysis. With the rapid growth of Deep Learning (DL) algorithms in image-based applications, CAD systems can now be trained by DL to provide more advanced capability (i.e., the capability of artificial intelligence [AI]) to best assist clinicians).

Description:

The mandibular third molar extraction, considered one of the most common surgeries in oral and maxillofacial field, it can be associated with several postoperative complications, like pain, bleeding, swelling, and inferior alveolar nerve (IAN) injury or complete damage, impairing the quality of life of the affected patients. The incidence of temporary IAN injury caused by mandibular third molar extraction was 0.4-8.4%, while the incidence of permanent injury is less than 1% [1, 2]. However, due to the high occurrence of impacted mandibular third molar, a large number of patients suffer from IAN injury caused by impacted mandibular third molar extraction [3]. The most significant risk factor of IAN injury caused by mandibular third molar extraction is the proximity of the root of the mandibular third molar to the mandibular canal [1, 2, 4, 5]. So, comprehensive preoperative analysis and evaluation of the anatomical structures are essential before impacted mandibular third molar extraction to decrease the IAN injury risk.

The panoramic radiography is not that much accurate in displaying the relation between impacted mandibular third molar extraction and IAN due to the superimposition and inherent limitations. The accuracy of predicting the probability the (IAN) injury during the impacted mandibular third molar extraction using panoramic radiographs were controversial [6].

Cone beam computed tomography (CBCT), A (3D) imaging modality, provides accurate 3D information with decreased radiation dose than medical CT [7]. It was demonstrated that CBCT was a better and accurate radiographic method than panoramic radiography for evaluating the relationship between mandibular third molar and (IAN) [6, 8]. So that, CBCT has been considered as the modality of choice for preoperative assessment of complicated mandibular third molar extraction [9].

Deep learning, one of artificial intelligence subsets, had a rapid progression and has a significant role in medical fields. One of the deep learning models, guided learning of the convolutional neural network (CNN) is recently investigated, which has been proven to surpass human judgmental level in many medical imaging fields [12, 13]. After CNN was introduced to the maxillofacial field, it was used for the assessment, detection, categorization, and segmentation of the surrounding anatomical structures [14-18 Recently, deep learning based on CNN models has been used for the impacted mandibular third molar and mandibular canal detection and segmentation on panoramic radiographs and CBCT [15, 18, 30], the classification and staging of development [31, 32], and the approximation measurements of the impacted mandibular third molar on panoramic radiographs [33]. Fukuda et al. compared 3 CNNs for classification of the impacted mandibular third molar and mandibular canal relation with panoramic radiographs [34]. Yoo et al. proposed a CNN-based approach to assess the stalemate of the impacted mandibular third molar extraction using panoramic radiographs [35]. So, as mentioned before, panoramic radiography can't accurately describe the anatomical structures due to the superimposition that happens in the (2D) imaging modalities. Orhan et al. reported an AI application (Diagnocat, Inc.) based on CNN with high precision in detecting the M3 and assessment of the number of roots related to adjacent anatomical structures

Recruitment information / eligibility

• Status: Recruiting
• Enrollment: 50
• Est. completion date: December 2023
• Est. primary completion date: December 2023
• Gender: All
• Age group: 25 Years to 65 Years

Eligibility

Inclusion Criteria:

• • CBCT Scans showing Mandibular third molar of patients aging from 25 to 65 years old

• The FOV should clearly show the third molar completely with its roots and the IAN.

• Voxel size of 0.2mm.

• Mandibular third molars. Absence of artifacts, dental implants in the adjacent teeth.

Exclusion Criteria:

• • CBCT images of sub-optimal quality or artifacts/high scatter interfering with proper assessment.

Related Conditions & MeSH terms

• Artificial Intelligence

Intervention

Diagnostic Test:
• CNN based model
It is an automatic detector model based on convolution neural network created by computer science expert

Locations

Country	Name	City	State
Egypt	Faculty of dentistry cairo university	Cairo

Sponsors (1)

Lead Sponsor	Collaborator
Cairo University

Country where clinical trial is conducted

Egypt, 

References & Publications (6)

Ghaeminia H, Meijer GJ, Soehardi A, Borstlap WA, Mulder J, Bergé SJ. Position of the impacted third molar in relation to the mandibular canal. Diagnostic accuracy of cone beam computed tomography compared with panoramic radiography. Int J Oral Maxillofac Surg. 2009 Sep;38(9):964-71. doi: 10.1016/j.ijom.2009.06.007. Epub 2009 Jul 28. — View Citation

Gülicher D, Gerlach KL. Sensory impairment of the lingual and inferior alveolar nerves following removal of impacted mandibular third molars. Int J Oral Maxillofac Surg. 2001 Aug;30(4):306-12. — View Citation

Kim JW, Cha IH, Kim SJ, Kim MR. Which risk factors are associated with neurosensory deficits of inferior alveolar nerve after mandibular third molar extraction? J Oral Maxillofac Surg. 2012 Nov;70(11):2508-14. doi: 10.1016/j.joms.2012.06.004. Epub 2012 Aug 15. — View Citation

Kwak GH, Kwak EJ, Song JM, Park HR, Jung YH, Cho BH, Hui P, Hwang JJ. Automatic mandibular canal detection using a deep convolutional neural network. Sci Rep. 2020 Mar 31;10(1):5711. doi: 10.1038/s41598-020-62586-8. — View Citation

Leung YY, Cheung LK. Risk factors of neurosensory deficits in lower third molar surgery: an literature review of prospective studies. Int J Oral Maxillofac Surg. 2011 Jan;40(1):1-10. doi: 10.1016/j.ijom.2010.09.005. Epub 2010 Oct 28. Review. — View Citation

Tay AB, Go WS. Effect of exposed inferior alveolar neurovascular bundle during surgical removal of impacted lower third molars. J Oral Maxillofac Surg. 2004 May;62(5):592-600. — View Citation

Outcome

Type	Measure	Description	Time frame	Safety issue
Primary	Accuracy of the automatic evaluation of the relationship between mandibular third molar and the mandibular canal.	Accuracy of the deep learning model in automatic evaluation of mandibular third molar teeth and mandibular canal relationship.	baseline