relu

Link to paper

Influence of dental fillings and tooth type on the performance of a novel artificial intelligence-driven tool for automatic tooth segmentation on CBCT images – A validation study

‍

Rocharles Cavalcante Fontenele ^ab, Maurício do Nascimento Gerhardt ^ac, Jáder Camilo Pinto ^ad, Adriaan Van Gerven ^e, Holger Willems ^e, Reinhilde Jacobs ^a ^f ^g, Deborah Queiroz Freitas ^b

^aOMFS IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Leuven, Belgium

^bDepartment of Oral Diagnosis, University of Campinas, Piracicaba, Brazil

‍^cDepartment of Prosthodontics, Pontifical Catholic University of Rio Grande do Sul, Porto Alegre, Brazil

‍^dDepartment of Restorative Dentistry, São Paulo State University, School of Dentistry, Araraquara, Brazil

‍^eRelu, Innovatie-en incubatiecentrum KU Leuven, Leuven, Belgium

‍^fDepartment of Oral and Maxillofacial Surgery, University Hospitals Leuven, Leuven, Belgium

‍^gDepartment of Dental Medicine, Karolinska Institute, Stockholm, Sweden

‍

Abstract

Objectives

To assess the influence of dental fillings on the performance of an artificial intelligence (AI)-driven tool for tooth segmentation on cone-beam computed tomography (CBCT) according to the type of tooth.

Methods

A total of 175 CBCT scans (500 teeth) were recruited for performing training (140 CBCT scans - 400 teeth) and validation (35 CBCT scans - 100 teeth) of the AI convolutional neural networks. The test dataset involved 74 CBCT scans (226 teeth), which was further divided into control and experimental groups depending on the presence of dental filling: without filling (control group: 24 CBCT scans – 113 teeth) and with coronal and/or root filling (experimental group: 50 CBCT scans – 113 teeth). The segmentation performance for both groups was assessed. Additionally, 10% of each tooth type (anterior, premolar, and molar) was randomly selected for time analysis according to manual, AI-based and refined-AI segmentation methods.

Results

The presence of fillings significantly influenced the segmentation performance (p<0.05). However, the accuracy metrics showed an excellent range of values for both control (95% Hausdorff Distance (95% HD): 0.01–0.08 mm; Intersection over union (IoU): 0.97–0.99; Dice similarity coefficient (DSC): 0.98–0.99; Precision: 1.00; Recall: 0.97–0.99; Accuracy: 1.00) and experimental groups (95% HD: 0.17–0.25 mm; IoU: 0.91–0.95; DSC: 0.95–0.97; Precision:1.00; Recall: 0.91–0.95; Accuracy: 0.99–1.00). The time analysis showed that the AI-based segmentation was significantly faster with a mean time of 29.8 s (p<0.001).

Conclusions

The proposed AI-driven tool allowed an accurate and time-efficient approach for the segmentation of teeth on CBCT images irrespective of the presence of high-density dental filling material and the type of tooth.

Clinical significance

Tooth segmentation is a challenging and time-consuming task, mainly in the presence of artifacts generated by dental filling material. The proposed AI-driven tool could offer a clinically acceptable approach for tooth segmentation, to be applied in the digital dental workflows considering its time efficiency and high accuracy regardless of the presence of dental fillings.

Introduction

The conventional dentistry workflows are being constantly replaced by digital workflows due to the incorporation of several computer-controlled tools, such as cone-beam computed tomography (CBCT), computer-aided-design/computer-assisted-manufacturing (CAD/CAM), three-dimensional (3D) printing, dynamic navigation and other advanced prototyping methodologies. All disciplines of dentistry have benefitted from these technological advancements and improved the efficiency of the oral healthcare system by offering a precise diagnosis, patient-specific treatment planning and follow-up evaluation [1].

One of the most critical steps in the digital dental workflows is the segmentation of teeth on CBCT images, which is a prerequisite for ensuring a correct diagnosis of tooth-related diseases and accurate virtual treatment planning [2], [3], [4], [5]. This step is most commonly performed by semi- or fully-automatic processes that require selecting a thresholding of gray values of the voxels that better encompass the region of interest [6,7]. However, these threshold-based approaches are prone to certain limitations which might significantly impact the accuracy of tooth segmentation, such as inability to distinguish tooth root from alveolar bone due to the presence of similar intensity profile, sensitivity to image noise, human performance variability depending on the operator's expertise and requirement of manual corrections which can be laborious and time-consuming. Furthermore, gray values thresholding is unreliable in the presence of artifacts generated by high-density dental materials, such as coronal fillings, metal posts and root fillings, due to higher variability of gray values at the region closest to the artifact generating material [4,7,8].

Recent advancements in the field of artificial intelligence (AI) have allowed the inclusion of innovative tools in the digital dental practice to overcome the inherent limitations associated with the classical tooth segmentation techniques [1,[2], [3], [4], [5]]. The availability of modern data-intensive computational resources and data availability have heavily contributed to developing robust deep learning models through the application of artificial neural networks inspired by the biological neural network of a human brain [9,10]. Convolutional neural network (CNN), a robust type of deep learning algorithm, has shown significant potential for the automated segmentation of pharyngeal airway [11], mandibular bone [9], and teeth [2,4,5,12,13].

Previous studies have reported the use of CNNs for performing automatic individual tooth segmentation from CBCT images with high accuracy [4,5,12,13]. However, a lack of evidence exists concerning the performance of these CNNs for the automated segmentation of teeth restored with coronal and/or root fillings. Thus, the influence of artifacts generated by high-density restorative materials on the performance of CNNs should be considered as the majority of patients do not offer a pristine dentition. Furthermore, the time required for performing segmentation has also received less attention, which is a key deciding factor for the clinical applicability of these networks.

Therefore, the aim of the present study was to assess the influence of dental fillings and the type of tooth on the performance of a CNN-based tool for automatic tooth segmentation on CBCT images.

Section snippets

Materials and methods

This study was conducted in compliance with the World Medical Association Declaration of Helsinki on medical research. Ethical approval was obtained from the Local Institutional Ethics Board (reference number: B322201525552). Informed consent was not required as patient-specific information was anonymized.

Results

The manual segmentation demonstrated optimal values for all accuracy metrics, thereby indicating that the expert was optimally calibrated (95% HD - 0.27 mm, IoU – 0.92, DSC – 0.96; Precision – 0.95, Recall – 0.96, and Accuracy – 1.00). The ICC for the time required to perform both manual segmentation and refinements was 0.97. In addition, the intra-examiner agreements for the need for refinements on the automatic segmentation (weighted Kappa=0.92) and the qualitative visual estimation of the

Discussion

Based on the hypothesis that the presence of restorative material artifacts could reduce the quality of the automated segmentation performed by an AI-driven tool, the present study was conducted to validate an innovative AI-driven tool for an accurate and time-efficient automated segmentation of teeth with and without dental filling material. Our results showed that the present AI-driven tool tested showed high accuracy and fast performance to generate 3D tooth models, even in the presence of

Conclusions

The dental filling materials influenced the AI-driven tool performance, mainly for the anterior teeth. However, the AI-driven tool proposed showed high accuracy metrics and a time-efficient approach to provide 3D tooth models from CBCT images regardless of artifacts generated by these high-density materials and the type of tooth.

Credit authorship contribution statement

Rocharles Cavalcante Fontenele: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing – original draft. Maurício do Nascimento Gerhardt: Conceptualization, Methodology, Validation, Software, Writing – original draft, Writing – review & editing. Jáder Camilo Pinto: Formal analysis, Investigation, Writing – review & editing. Adriaan Van Gerven: Conceptualization, Methodology, Software, Validation, Writing – review & editing. Holger Willems: Conceptualization,

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001.

References (23)

P. Lahoud et al. Artificial intelligence for fast and accurate 3-dimensional tooth segmentation on cone-beam computed tomography J. Endod. (2021)

E. Shaheen et al. A novel deep learning system for multi-class tooth segmentation and classification on cone beam computed tomography. A validation study J. Dent. (2021)

P.J. Verhelst et al. Layered deep learning for automatic mandibular segmentation in cone-beam computed tomography J. Dent. (2021)

S. Shujaat et al. Automated segmentation of the pharyngeal airway space with convolutional neural network J. Dent. (2021)

M. EzEldeen et al. 3-dimensional analysis of regenerative endodontic treatment outcome J. Endod. (2015)

S. Shujaat et al. Integration of imaging modalities in digital dental workflows - possibilities, limitations, and potential future developments Dentomaxillofac. Radiol. (2021)

S. Lee et al. Automated CNN-Based tooth segmentation in cone-beam CT for dental implant planning IEEE Access (2020)

A.F. Leite et al. Artificial intelligence-driven novel tool for tooth detection and segmentation on panoramic radiographs Clin. Oral Investig. (2021)

B. Hassan et al. Van der Stelt, Influence of scanning and reconstruction parameters on quality of three-dimensional surface models of the dental arches from cone beam computed tomography Clin. Oral Investig. (2010)