Layered deep learning for automatic mandibular segmentation in cone-beam computed tomography

Verhelst, Pieter-Jan; Smolders, A.; Beznik, Thomas; Meewis, J. M. J. M.; Vandemeulebroucke, Arne; Shaheen, Eman; Gerven, Adriaan Van; Willems, Holger; Politis, Constantinus; Jacobs, Reinhilde

doi:10.1016/j.jdent.2021.103786

Cited by 95 publications

(52 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This is the first study to our knowledge to use the new 3D UNETR architecture to segment multiple anatomic skeletal, dental, and soft tissue structures in the craniofacial complex of CBCT scans. Recent studies have focused on only one specific facial structure such as the maxilla, [ 17 ], mandible [ 18 ] or airway [ 12 ], and used smaller samples from a single CBCT acquisition protocol, thus, those algorithms are not yet generalizable like the proposed AMASSS.…”

Section: Discussionmentioning

confidence: 99%

Automatic multi-anatomical skull structure segmentation of cone-beam computed tomography scans using 3D UNETR

Gillot

Baquero

et al. 2022

PLoS ONE

View full text Add to dashboard Cite

The segmentation of medical and dental images is a fundamental step in automated clinical decision support systems. It supports the entire clinical workflow from diagnosis, therapy planning, intervention, and follow-up. In this paper, we propose a novel tool to accurately process a full-face segmentation in about 5 minutes that would otherwise require an average of 7h of manual work by experienced clinicians. This work focuses on the integration of the state-of-the-art UNEt TRansformers (UNETR) of the Medical Open Network for Artificial Intelligence (MONAI) framework. We trained and tested our models using 618 de-identified Cone-Beam Computed Tomography (CBCT) volumetric images of the head acquired with several parameters from different centers for a generalized clinical application. Our results on a 5-fold cross-validation showed high accuracy and robustness with a Dice score up to 0.962±0.02. Our code is available on our public GitHub repository.

show abstract

Section: Discussionmentioning

confidence: 99%

Automatic multi-anatomical skull structure segmentation of cone-beam computed tomography scans using 3D UNETR

Gillot

Baquero

et al. 2022

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…Because CBCT data has an intrinsic low image contrast, lack of Hounsfield units, and increased noise and artifacts compared to multi-slice CT, semi-automatic segmentation requires manual edits which could influence the reliability. Further developments with regard to fully automatic segmentation with artificial intelligence (AI) models could solve this shortcoming [ 30 , 31 ].…”

Section: Discussionmentioning

confidence: 99%

Effect of voxel size in cone-beam computed tomography on surface area measurements of dehiscences and fenestrations in the lower anterior buccal region

et al. 2022

View full text Add to dashboard Cite

Objectives This study aims to assess whether different voxel sizes in cone-beam computed tomography (CBCT) affected surface area measurements of dehiscences and fenestrations in the mandibular anterior buccal region. Materials and methods Nineteen dry human mandibles were scanned with a surface scanner (SS). Wax was attached to the mandibles as a soft tissue equivalent. Three-dimensional digital models were generated with a CBCT unit, with voxel sizes of 0.200 mm (VS200), 0.400 mm (VS400), and 0.600 mm (VS600). The buccal surface areas of the six anterior teeth were measured (in mm2) to evaluate areas of dehiscences and fenestrations. Differences between the CBCT and SS measurements were determined in a linear mixed model analysis. Results The mean surface area per tooth was 88.3 ± 24.0 mm2, with the SS, and 94.6 ± 26.5 (VS200), 95.1 ± 27.3 (VS400), and 96.0 ± 26.5 (VS600), with CBCT scans. Larger surface areas resulted in larger differences between CBCT and SS measurements (− 0.1 β, SE = 0.02, p < 0.001). Deviations from SS measurements were larger with VS600, compared to VS200 (1.3 β, SE = 0.05, P = 0.009). Fenestrations were undetectable with CBCT. Conclusions CBCT imaging magnified the surface area of dehiscences in the anterior buccal region of the mandible by 7 to 9%. The larger the voxel size, the larger the deviation from SS measurements. Fenestrations were not detectable with CBCT. Clinical relevance CBCT is an acceptable tool for measuring dehiscences but not fenestrations. However, CBCT overestimates the size of dehiscences, and the degree of overestimation depends on the actual dehiscence size and CBCT voxel size employed.

show abstract

“…The mean duration of the segmentation ranged from 6.89 to 83 min in the different studies. Many studies are currently focused on automatizing the segmentation of bone, and with the further development of machine learning for these purposes, efficient segmentation methods will hopefully continue to improve (Liu et al, 2021;Verhelst et al, 2021;Deng et al, 2022). However, especially the fracture fragment separation is labor-intensive and harder to automatize.…”

Section: Discussionmentioning

confidence: 99%

Computer-assisted preoperative planning of bone fracture fixation surgery: A state-of-the-art review

Moolenaar

Tümer

Checa

2022

Front. Bioeng. Biotechnol.

View full text Add to dashboard Cite

Background: Bone fracture fixation surgery is one of the most commonly performed surgical procedures in the orthopedic field. However, fracture healing complications occur frequently, and the choice of the most optimal surgical approach often remains challenging. In the last years, computational tools have been developed with the aim to assist preoperative planning procedures of bone fracture fixation surgery.Objectives: The aims of this review are 1) to provide a comprehensive overview of the state-of-the-art in computer-assisted preoperative planning of bone fracture fixation surgery, 2) to assess the clinical feasibility of the existing virtual planning approaches, and 3) to assess their clinical efficacy in terms of clinical outcomes as compared to conventional planning methods.Methods: A literature search was performed in the MEDLINE-PubMed, Ovid-EMBASE, Ovid-EMCARE, Web of Science, and Cochrane libraries to identify articles reporting on the clinical use of computer-assisted preoperative planning of bone fracture fixation.Results: 79 articles were included to provide an overview of the state-of-the art in virtual planning. While patient-specific geometrical model construction, virtual bone fracture reduction, and virtual fixation planning are routinely applied in virtual planning, biomechanical analysis is rarely included in the planning framework. 21 of the included studies were used to assess the feasibility and efficacy of computer-assisted planning methods. The reported total mean planning duration ranged from 22 to 258 min in different studies. Computer-assisted planning resulted in reduced operation time (Standardized Mean Difference (SMD): -2.19; 95% Confidence Interval (CI): -2.87, -1.50), less blood loss (SMD: -1.99; 95% CI: -2.75, -1.24), decreased frequency of fluoroscopy (SMD: -2.18; 95% CI: -2.74, -1.61), shortened fracture healing times (SMD: -0.51; 95% CI: -0.97, -0.05) and less postoperative complications (Risk Ratio (RR): 0.64, 95% CI: 0.46, 0.90). No significant differences were found in hospitalization duration. Some studies reported improvements in reduction quality and functional outcomes but these results were not pooled for meta-analysis, since the reported outcome measures were too heterogeneous.Conclusion: Current computer-assisted planning approaches are feasible to be used in clinical practice and have been shown to improve clinical outcomes. Including biomechanical analysis into the framework has the potential to further improve clinical outcome.

show abstract

Layered deep learning for automatic mandibular segmentation in cone-beam computed tomography

Cited by 95 publications

References 34 publications

Automatic multi-anatomical skull structure segmentation of cone-beam computed tomography scans using 3D UNETR

Automatic multi-anatomical skull structure segmentation of cone-beam computed tomography scans using 3D UNETR

Effect of voxel size in cone-beam computed tomography on surface area measurements of dehiscences and fenestrations in the lower anterior buccal region

Computer-assisted preoperative planning of bone fracture fixation surgery: A state-of-the-art review

Contact Info

Product

Resources

About