Journal of Dentistry 143 (2024) 104862

Contents lists available at ScienceDirect

Journal of Dentistry
journal homepage: www.elsevier.com/locate/jdent

Review article

Artificial intelligence serving pre-surgical digital implant planning: A

scoping review
Bahaaeldeen M. Elgarba a, *, Rocharles Cavalcante Fontenele b, Mihai Tarce c, Reinhilde Jacobs d
a
OMFS IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven & Department of Oral and Maxillofacial Surgery, University
Hospitals, Campus Sint-Rafael, 3000 Leuven, Belgium & Department of Prosthodontics, Faculty of Dentistry, Tanta University, 31511 Tanta, Egypt
b
OMFS IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven & Department of Oral and Maxillofacial Surgery, University
Hospitals, Campus Sint-Rafael, 3000 Leuven, Belgium
c
Division of Periodontology & Implant Dentistry, Faculty of Dentistry, The University of Hong Kong, Hong Kong SAR, China & Periodontology and Oral Microbiology,
Department of Oral Health Sciences, Faculty of Medicine, KU Leuven, Leuven, Belgium
d
OMFS IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven & Department of Oral and Maxillofacial Surgery, University
Hospitals, Campus Sint-Rafael, 3000 Leuven, Belgium & Department of Dental Medicine, Karolinska Institute, Stockholm, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: Objectives: To conduct a scoping review focusing on artificial intelligence (AI) applications in presurgical dental
Artificial intelligence implant planning. Additionally, to assess the automation degree of clinically available pre-surgical implant
Deep learning planning software.
Dental implant
Data and sources: A systematic electronic literature search was performed in five databases (PubMed, Embase,
Implant dentistry
3D imaging
Web of Science, Cochrane Library, and Scopus), along with exploring gray literature web-based resources until
Cone-beam computed tomography November 2023. English-language studies on AI-driven tools for digital implant planning were included based on
an independent evaluation by two reviewers. An assessment of automation steps in dental implant planning
software available on the market up to November 2023 was also performed.
Study selection and results: From an initial 1,732 studies, 47 met eligibility criteria. Within this subset, 39 studies
focused on AI networks for anatomical landmark-based segmentation, creating virtual patients. Eight studies
were dedicated to AI networks for virtual implant placement. Additionally, a total of 12 commonly available
implant planning software applications were identified and assessed for their level of automation in pre-surgical
digital implant workflows. Notably, only six of these featured at least one fully automated step in the planning
software, with none possessing a fully automated implant planning protocol.
Conclusions: AI plays a crucial role in achieving accurate, time-efficient, and consistent segmentation of
anatomical landmarks, serving the process of virtual patient creation. Additionally, currently available systems
for virtual implant placement demonstrate different degrees of automation. It is important to highlight that, as of
now, full automation of this process has not been documented nor scientifically validated.
Clinical significance: Scientific and clinical validation of AI applications for presurgical dental implant planning
is currently scarce. The present review allows the clinician to identify AI-based automation in presurgical dental
implant planning and assess the potential underlying scientific validation.

1. Introduction These encompass digital patient data acquisition, three-dimensional

(3D) anatomical landmark visualization, digital treatment planning,
Digital workflows in implant dentistry offer substantial benefits, and the design of final prostheses [2]. A variety of devices, including
enhancing treatment predictability [1]. In this context, the digitalization cone-beam computed tomography (CBCT) devices, intraoral scanners
of dental implant planning workflows plays an important role in inte- (IOS), facial scans, computer-aided-design/computer-assisted-
grating different digital technologies across various treatment stages. manufacturing (CAD/CAM), and implant planning software programs,

* Corresponding author at: OMFS IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Kapucijnenvoer 7, BE-3000
Leuven, Belgium.
E-mail address: bahaa.elgarba@kuleuven.be (B.M. Elgarba).

https://doi.org/10.1016/j.jdent.2024.104862
Received 14 December 2023; Received in revised form 22 January 2024; Accepted 24 January 2024
Available online 8 February 2024
0300-5712/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

are commercially available to facilitate the integration of digital work- Table 1

flows for planning in implant dentistry [3,4]. Consequently, clinicians Study search strategy within each database.
are required to make informed decisions regarding digital methods, Databases Search strategy
taking into account treatment complexity, time efficiency, and cost
PubMed (“cone-beam computed tomography”[Mesh] OR “cone-beam computed
analysis [5]. tomography”[tiab] OR “CBCT”[tiab] OR “imaging, three-
An efficient digital workflow for implant dentistry requires hardware dimensional”[tiab] OR “mandibular nerve”[Mesh] OR “mandibular
for image acquisition and software for multimodal image integration nerve”[tiab] OR “jaw”[tiab] OR “dental prosthesis”[tiab] OR “oral
and treatment planning [1]. Standard image acquisition techniques for implant*” [tiab] OR ‘’tooth implant*’’[tiab] OR ‘’teeth implant*’’[tiab]
OR ‘’endosseous implant*’’[tiab]) AND (“artificial intelligence”[MeSH]
implant planning include CBCT, IOS and facial scanning [6,7]. Addi- OR “artificial intelligence”[tiab] OR “AI”[tiab] OR “computational
tionally, software is crucial for multimodal image integration, compre- intelligence”[tiab] OR “machine intelligence”[tiab] OR “computer
hensive treatment planning, and image analysis [8]. These facilities not reasoning”[tiab] OR “computer vision system*”[tiab] OR “knowledge
only streamline the digital implant process for uncomplicated cases but acquisition”[tiab] OR “knowledge representation*”[tiab] OR “expert
systems”[tiab] OR “natural Language Processing”[tiab] OR “neural
also enable accurate anatomical structure segmentation. This, in turn,
Network*”[tiab] OR “ambient intelligence”[tiab] OR “automated
enables customized implant planning procedures for individual patients, reasoning”[tiab] OR “computer heuristics”[tiab] OR “learning
including the utilization of customized dental implants, personalized algorithm*”[tiab] OR “reinforcement learning”[tiab])
titanium meshes in guided bone regeneration, and the implementation Embase ‘cone beam computed tomography’/exp OR ‘cone beam computed
of CAD/CAM scaffolds for bone augmentation [9–12]. tomography’:ti,ab,kw OR ‘CBCT’:ti,ab,kw OR ‘imaging, three-
dimensional’:ti,ab,kw OR ‘mandibular nerve’:ti,ab,kw OR ‘jaw’:ti,
Nevertheless, the multiple steps involved in image acquisition and ab,kw ‘tooth implant*’:ti,ab,kw OR ‘teeth implant*’:ti,ab,kw OR
software manipulations render implant planning a complex, time- ‘endosseous implant*’:ti,ab,kw OR ‘dental prosthesis’:ti,ab,kw OR
consuming, and user-dependent process, requiring a high level of ‘dental implant*’:ti,ab,kw AND ‘artificial intelligence’/exp OR
training and expertise [13]. To streamline this workflow and simplify ‘artificial intelligence’:ti,ab,kw OR ‘ambient intelligence’:ti,ab,kw
OR ‘automated reasoning’:ti,ab,kw Or ‘computer heuristics’:ti,ab,kw
the digital treatment steps, artificial intelligence (AI) techniques have
OR ‘machine* intelligence’:ti,ab,kw OR ‘artificial neural network’:ti,
been proposed [14]. AI comprises computer systems’ capability to ab,kw OR ‘machine learning’:ti,ab,kw OR ‘learning algorithm*’:ti,
perform tasks similar to those performed by humans. From this concept, ab,kw OR ‘reinforcement learning’:ti,ab,kw OR ‘computational
deep learning emerges as a method grounded in neural network intelligence’:ti,ab,kw OR ‘computer reasoning’:ti,ab,kw OR ‘AI-
learning, employing artificial neural networks to capture features within based’:ti,ab,kw OR ‘computer vision System*’:ti,ab,kw OR
‘knowledge acquisition’:ti,ab,kw OR ‘knowledge representation*’:ti,
complex data sets [15]. The significance of AI in dentistry lies in its
ab,kw OR ‘deep learning’:ti,ab,kw OR ‘supervised machine learning’:
potential to enhance accuracy, efficiency, and patient outcomes, thereby ti,ab,kw OR ‘unsupervised machine learning’:ti,ab,kw OR ‘expert
revolutionizing the delivery and experience of dental care. Moreover, it systems’:ti,ab,kw OR ‘fuzzy Logic’:ti,ab,kw OR ‘natural language
facilitates the automation of repetitive tasks, reducing human error and processing’:ti,ab,kw OR ‘neural networks computer*’:ti,ab,kw OR
‘robotics’:ti,ab,kw
enhancing procedural efficiency, which, in turn, may reduce the risk of
Web of TS→(“cone beam computed tomography” OR “cone beam computed
complications and the assurance of successful outcomes [16–22]. Science tomography” OR “CBCT” OR “imaging, three-dimensional” OR
Previous reviews have highlighted the potential of artificial intelli- Core “mandibular nerve” OR “jaw” “tooth implant*” OR “teeth implant*”
gence (AI) models in various aspects of implant dentistry and post- collection OR “endosseous implant*” OR “dental prosthesis” OR “dental
treatment monitoring. This encompasses applications such as dental implant*”) AND TS→(“artificial intelligence” OR “ambient
intelligence” OR “automated reasoning” OR “computer heuristics”
implant recognition, peri-implant bone loss detection, and implant
OR “machine* intelligence” OR “artificial neural network” OR
failure determination, demonstrating the capability for AI to contribute “machine learning” OR “learning algorithm*” OR “reinforcement
to enhance diagnostic and treatment monitoring capabilities in implant learning” OR “computational intelligence” OR “computer reasoning”
dentistry by evaluating bone dimensions and identifying the positions OR “AI-based” OR “computer vision System*” OR “knowledge
acquisition” OR “knowledge representation*” OR “deep learning”
for potential dental implant placement on CBCT images [14,23,24].
OR “supervised machine learning” OR “unsupervised machine
While AI holds promise for these tasks, its overall accuracy may vary due learning” OR “expert systems” OR “fuzzy Logic” OR “natural
to the extensive array of digital technologies employed in the digital language processing” OR “neural networks computer*” OR
workflow [1]. To the best of our knowledge, the scientific literature “robotics”)
lacks evidence regarding the interpretation of AI in the context of digital Cochrane [mh “cone-beam computed tomography”] or [mh “imaging, three-
dimensional”] or (“cone-beam computed tomography” OR “CBCT”
presurgical implant planning. Thus, conducting an individual analysis of
OR “imaging, three-dimensional” OR “mandibular nerve” OR “jaw”
different treatment steps can contribute to substantiating the use of AI in OR “dental prosthesis” OR “oral implant*” OR ‘’tooth implant*’’ OR
dental implant planning. Nonetheless, it remains unclear which specific ‘’teeth implant*’’ OR ‘’endosseous implant*’’):ti,ab,kw AND [mh
steps within the digital implant planning workflow are automated and “artificial intelligence”] OR (“artificial intelligence” OR “AI” OR
“computational intelligence” OR “machine intelligence” OR
how AI can improve dental implant treatment planning.
“computer reasoning” OR “computer vision system*” OR
This scoping review has two main objectives. The primary aim is to “knowledge acquisition” OR “knowledge representation*” OR
explore the existing literature and provide an updated overview of the “expert systems” OR “natural Language Processing” OR “neural
leading AI applications in digital dental implant planning. Secondly, this Network*” OR “ambient intelligence” OR “automated reasoning” OR
review aims to assess available dental implant planning software and “computer heuristics” OR “learning algorithm*” OR “reinforcement
learning”)
examine their level of automation. Based on these aims, the following
Scopus (TITLE-ABS (“cone beam computed tomography” OR “cone beam
focused questions were addressed: 1) What digital dental implant computed tomography” OR “CBCT” OR “imaging, three-
workflow steps are already automated? 2) How does this automation dimensional” OR “mandibular nerve” OR “jaw” “tooth implant*” OR
enhance digital dental implant planning, particularly in terms of accu- “teeth implant*” OR “endosseous implant*” OR “dental prosthesis”
racy, time efficiency, and consistency? 3) Which specific automations OR “dental implant*”)) AND (TITLE-ABS (“artificial intelligence” OR
“deep learning” OR “AI-based” OR “machine learning” OR “artificial
within the different clinical applications related to dental implant neural network” OR “ambient intelligence” OR “automated
treatment planning are based on AI tools? reasoning” OR “learning algorithm*” OR “computer vision System*”
OR “knowledge acquisition” OR “knowledge representation*” OR
2. Materials and methods “deep learning” OR “supervised machine learning” OR
“unsupervised machine learning” OR “computer heuristics”))

This scoping review was based on a systematic literature review

protocol for literature search, following the guidelines outlined in the
Preferred Reporting Items for Systematic Reviews and Meta-Analyses

2
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

Fig. 1. Steps of digital workflow in dental implant planning.

(PRISMA) statement [25]. Furthermore, this protocol was registered in ↑ Population/problem (P) → Digital dental implant workflow
the International Prospective Register of Systematic Reviews (PROS- ↑ Intervention (I) → Artificial intelligence (AI)
PERO) database under the registration number CRD42022295683. ↑ Comparison I → Human Intelligence (HI)
The research question and, consequently, the search strategy were ↑ Outcome (O) → Accuracy, time efficiency, and consistency
formulated using the PICOS acronym as follows: ↑ Study design (S) → Retrospective and prospective studies

Fig. 2. PRISMA (2020) flowchart regarding the study selection.

3
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

Fig. 3. Distribution of the included studies over the last 5 years.

2.1. Electronic searches information was either sourced from existing materials, if available or
obtained by reaching out to the authors through email. If the two re-
The electronic search was conducted in five databases (PubMed, viewers could not reach a consensus on the final critical appraisal
Embase, Web of Science, Cochrane, and Scopus libraries) until through discussion, a third reviewer (RJ) was involved. The third re-
November 2023, each implementing a search strategy tailored to its viewer’s role was to provide an additional evaluation and help in
platform, as seen in Table 1. Furthermore, a manual search in gray resolving any disagreements.
literature databases, customized Google search engines, targeted web-
sites, and reference/citation lists of included studies was conducted to
ensure the comprehensive inclusion of all relevant studies. All titles and 2.4. Available Dental Implant Planning Programs
abstracts were reviewed during the screening process to include studies
that investigated AI-based tools employed in digital dental implant Two reviewers (BE and MT) conducted extensive screening of
planning with a focus on 3D imaging modalities. Studies that were not available implant planning programs, including open-source and avail-
original or not in English or were conducted solely in vitro were excluded able license programs, up to November 2023 to assess the level of
from this study. automation in implant planning programs available for practical use in
accordance with the digital dental implant workflow, as illustrated in
2.2. Selection of studies Fig. 1. One reviewer (BE) performed the primary screening, and the
results were subsequently double-checked by a second reviewer (MT).
The Rayyan platform (https://rayyan.qcri.org) was used to auto- The evaluation for each software program included assessing the type of
matically eliminate duplicates and screen the titles and abstracts of the 3D modality utilized and the software’s level of automation (e.g.,
retrieved articles. Once the bibliography was uploaded, the platform manual, semi-automatic, or automated) within the different steps of the
independently screened the articles for further assessment by two re- digital workflow. These steps comprised anatomical landmark segmen-
viewers (BE and RCF). Articles from additional sources as mentioned tation, multimodal image registration, bone volume measurement,
before, could also be added for screening alongside the previously digital wax up, implant dimensions, implant location, and surgical guide
included articles. Finally, it summarized the duplicate, included, and design.
excluded articles at the end of the study selection process.
The full text of these potentially eligible articles, as well as those 3. Results
abstracts that did not provide sufficient information to allow decision-
making, were retrieved and assessed by two independent reviewers. In 3.1. Eligible Studies
addition, the reviewers cross-checked the decisions. Any differences
between the two reviewers were discussed and settled by consensus after Fig. 2 depicts the PRISMA flowchart, providing an overview of the
consulting a third expert (RJ). results obtained from the literature search. Out of the 1,732 articles
identified in the databases, 48 abstracts were selected for full-text
2.3. Data extraction analysis. After a thorough examination of the complete texts, 45 arti-
cles were found to meet the eligibility criteria [13,17–21,26–51,53,
One reviewer (BE) extracted data, and later, another reviewer (RCF) 55–66], while three studies were excluded. One study was excluded as it
independently verified all the extracted information. Data extraction was a technical report [67], and two others were excluded due to their
was done according to a prepared list consisting of the following study lack of relevance to the subject of this review [68,69] (Table 1 Supple-
information: objective, region of interest (i.e., anatomical region mentary). Additionally, two articles were included from a manual search
investigated), study dataset, results, AI network, program name, and on external websites (e.g., Google search engines) [52,54]. Thus, a total
conclusion. For any information not explicitly detailed in the article, of 47 original articles were incorporated into this review.

4
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

Table 2
Data extraction of included studies concerning to AI-based virtual implant patient creation.

(continued on next page)

5
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

Table 2 (continued )

3.2. Time frame of included studies shared the two aspects by providing results for AI-based anatomic seg-
mentation (e.g., mandibular canals, maxillary sinus, nasal fossae, and
Fig. 3 illustrates the distribution of studies across various years. The edentulous zone) and virtually automated measurement of bone height
included articles were published within the timeframe of 2019-2023. and thickness [44].
Moreover, the number of articles related to the subject of investigation All studies targeted at segmenting anatomical landmarks used an AI
has progressively increased over the years, with the largest number of model based on a 3D U-network, which is now the most commonly
articles published in 2022 (n→18) [18,19,21,28,29,32,34,42,43,45,46, employed convolutional neural network (CNN) in the medical field for
49–53,62,66]. image segmentation [70]. In contrast, studies focusing on virtual
implant placement showed a variation in the type of AI networks uti-
lized, including 3D U-network [61–64,66], 2D U-network [13,65], and
3.3. Descriptive analysis of the included studies
augmented reality [60].
In terms of the anatomical regions automatically segmented by AI,
Tables 2 and 3 present data extracted from all included studies,
researchers predominantly focused on the mandibular canal, as evi-
organized by the objective of each study, dataset number, primary
denced by the inclusion of 14 studies in this review [21,47,48–59].
outcomes achieved, AI network used, AI-based program used, and
Tooth segmentation emerged as the second most popular area of interest
finally, the study’s conclusion. The 47 articles included in the current
for researchers, with nine studies dedicated to this aspect [18,26–33].
review explore two main aspects of using AI in the digital dental implant
The multicenter study by Cui et al. [42] employed the highest sample
workflow: anatomical landmark segmentation for virtual patient crea-
size composed of 3D scans, utilizing 4,938 CBCT scans to validate the
tion (n→39) [13,17–21,26–59] and the potential integration of AI into
robustness and generalizability of an AI system for teeth and alveolar
virtual implant placement (n→8) [13,60–66]. However, only one article

6
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

Table 3
Characteristics of included studies for AI-based virtual implant placement.

bone delineation on CBCT. In contrast, the smallest sample size observed have shown that AI is remarkably consistent, reliably performing the
in the studies included in the current reviews was composed of 5 CBCT same task given the same inputs [19,42,45,58].
scans [54]. While only eight articles (Table 3) discussed virtual dental implant
Regarding the accuracy of the AI tools developed for virtual implant placement [13,44,60–66], among these, two studies highlighted the
patient creation through maxillofacial anatomical landmark segmenta- significance of AI in reducing planning time [55,60], and two other
tion, a notable range of accuracy from 58% to 99.7% was observed when studies demonstrated AI’s ability to do fully automatic registration be-
compared to manual and semi-automatic segmentation approaches, tween CBCT and IOS [60,61]. Despite these findings indicating limited
which serve as the clinical standard reference. Furthermore, the time automation steps, they offer promising insights for potential future
required for automated segmentation emerged as a significant clinical research projects aiming at achieving fully automated implant planning.
finding in some studies, with AI demonstrating significantly faster seg- Fig. 4 illustrates the current state and future expectations for AI-based
mentation of anatomical landmarks than traditional methods [17–19, digital dental implant planning based on the results obtained in this
21,30,31,36,39,42,47,55,58]. The AI-based segmentation time ranged review.
from 1.5 seconds to less than 5 minutes [23,35]. Additionally, studies

Fig. 4. Recent availability and future expectations of AI-based automation steps in digital implant planning.

7
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

Table 4
Studied software programs for implant planning, including manufacturer companies, countries of origin, versions, and codes.
Software Code Company Country Version

Atomica.ai A Atomica USA 4.0.0

BlueSkyPlan B Blue Sky Bio USA 4.11.2
Co-DiagnostiX C Dental Wings-Straumann Canada 10.7
Dentiq D Dentis USA 1.3.5
DTX Studio implant E Nobel Biocare Switzerland 3.6.6.1
DTX Studio clinic F Dexis USA 1.7.5
Exoplan G Exocad Germany 3.1 Rijeka
3Shape Implant Studio H 3Shape Denmark 1.7.33.1
Romexis I Planmeca Finland 6.0
SICAT J Sicat Germany 1.3
Simplant Pro K Dentsply Sirona Sweden 18.5
Smop L Swissmeda AG Switzerland 2.17.1

Table 5
Relevant data input and automated steps within each implant planning program.

8
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

3.4. Implant planning software heightened risk of damaging the mandibular nerve during implant
insertion [73]. On the other hand, tooth segmentation came in second
In total, 12 implant planning software programs were identified with nine studies [18,26–33]. This prominence can be attributed to the
based on their license availability or open-source license, Table 4 pre- critical role tooth segmentation plays in the digital workflow, and
sents an alphabetical compilation of all studied programs, including automated segmentation can enhance the reliability of segmentation
their respective development companies, country of origin, and version. and reduce processing time [74].
Table 5 provides an overview of relevant data input and the auto- Furthermore, it is noteworthy that only two investigations [17,35]
mation within each step of the traditional digital workflow. These steps included in this review delved into the capacity of AI for segmenting
include 3D image acquisition, landmark segmentation, multimodal maxillofacial structures in the presence of artifacts generated by
image registration, bone volume measurement, digital wax-up, implant high-density material. It is important to emphasize that artifacts
dimension selection, implant location, and surgical guide design. All the commonly encountered in CBCT images, including those caused by
programs included could import CBCT and IOS scans; additionally, the beam-hardening phenomena, can significantly impair image quality. As
DTX Studio implant, DTX Studio clinic, Exoplan, and Romexis programs a result, they pose challenges for AI segmentation processes [17,75].
could import a face scan in addition to CBCT and IOS scans. Expanding research efforts in this domain is imperative to enhance the
Furthermore, the majority of the included programs employ manual applicability of AI algorithms across a broader range of clinical sce-
or semi-automated approaches for performing implant planning, with a narios. Moreover, it is essential to recognize that the datasets employed
few demonstrating variable levels of automation at different steps of the in these studies were collected from specific devices and may not be
digital implant placement planning process. For example, BlueSkyPlan, readily extrapolated to different CBCT devices. Consequently, the need
Romexis, DTX Studio Clinic, and Co-DiagnostiX demonstrated an auto- for multicentre studies arises as a crucial step to increase the general-
mated AI-based approach to segmenting anatomic landmarks and izability of the AI models already developed.
registering CBCT and IOS, while Atomica.ai, BlueSkyPlan, Co- Currently, the implant planning process commences with the precise
DiagnostiX, DTX Studio Clinic, and Romexis showed an automated 3D modeling of anatomical landmarks, followed by the multimodal
CBCT and IOS registration. registration of multiple 3D models (e.g., CBCT, IOS, and face scan-based
Only two programs, Atomica.ai and DTX Studio Clinic, featured a models) [14,76]. The existing literature provides limited evidence
fully automated digital wax-up, while other programs utilized a variety concerning fully automated CBCT-IOS registration. Although some
of manual and semi-automated methods for the wax-up process. SICAT studies have utilized AI tools to identify similar landmarks among
stands out with its complete automation of surgical guide design. various datasets, none of these approaches rely on automated landmark
Finally, concerning the availability of cloud-based software, Smop was segmentation, and the registration was based on clusters of points
the only software among those included in the current review that had a matching the datasets; these points were determined by a mathematical
cloud-based option. analysis in CBCT and IOS [13,65]. Future research should be conducted
toward a more comprehensive investigation to establish a robust foun-
4. Discussion dation for the accuracy and effectiveness of AI-driven registration
methods for different 3D datasets.
Artificial intelligence has revolutionized the digital dental implant Previous studies have investigated the automated segmentation of
workflow, overcoming individual variability and providing consistent bone jaws, encompassing the mandible and the maxillofacial complex,
and time-efficient daily tasks [14,24]. As a consequence, this scoping including the crestal bone area [19,20,36–39]. Additionally, AI has
review aimed to provide an overview of the use of AI technologies in found applications in bone analysis, such as determining bone mineral
digital dental implant planning. Additionally, it was aimed to assist density levels, measuring bone height and thickness, and identifying the
dental practitioners and implantologists in understanding the impact of most suitable drilling protocol according to bone quality [44,62–64].
AI on presurgical implant planning from theoretical (i.e., reviewing the However, not enough research has yet been conducted that tackles the
existing literature) and practical considerations (i.e., review of the automated assessment of bone quality or quantity. This may be due to
available software and their level of automation); Thus, the results of the limitation of CBCT scans for evaluating bone quality considering the
this current review might support professionals in achieving accurate gray values variability of CBCT images; however, there are still other
diagnoses and executing virtual implant placements for daily clinical structural analysis methods that can evaluate bone quality from CBCT
cases involving implant oral rehabilitation. Furthermore, this review [77]. Consequently, further integration of AI into future studies is
provides a broad perspective on the digital steps that still require warranted to assist dental practitioners in evaluating alveolar bone
automation, making them suitable for future research. This advance- morphology during implant planning.
ment can make the planning process more accessible and less It is worth noting that none of the studies or software applications
complicated. examined in this review offered a fully automated implant planning
Virtual implant patient creation represents the initial and crucial step solution based on AI utilizing 3D images. However, previous research
in presurgical implant planning. This process demands the precise seg- has explored the potential of AI for specific steps of implant planning,
mentation of anatomical landmarks from 3D imaging data, demanding a such as proposing implant dimensions based on CBCT images [61].
significant amount of effort and expertise to comprehend each patient’s Furthermore, a proposed protocol involving 3D planning for dental
unique anatomy [7,71-72]. Studies included in this review demon- implant placement was identified by combining AI with augmented
strated that AI could provide accurate, time-efficient, and consistent reality through the acquisition of 3D data using CBCT and IOS [60].
landmark segmentation, generating a 3D model that can be utilized to While these studies have shown promising outcomes in utilizing AI for
design and guide the entire implant placement procedure [18,19,21,30, virtual implant placement planning, it is essential to emphasize the need
31,36,39,42,45,47,55,58]. However, there are still some limitations for clinical studies to validate the application of these protocols across a
associated with AI-based segmentation, such as the lack of generaliz- range of clinical scenarios.
ability of the data used for AI model training [19,45,47] and artifacts One goal of the present scoping review was to identify the level of
caused by high-density materials that can impact the quality of seg- automation present in the software already available for dental implant
mentation [17,29]. Therefore, further studies can be conducted to placement planning. Notably, while a handful of programs incorporate
overcome these limitations. partial automation features, none offer a fully automated workflow.
Mandibular canal segmentation stands out with the highest number Automation was evident in the segmentation of dentomaxillofacial
of studies included, comprising 14 out of the 39 studies related to seg- structures and CBCT-IOS registration, and there was some indication of
mentation [21,47,48–59]. This prominence can be attributed to the it in certain planning steps, such as digital wax-up and surgical guide

9
B.M. Elgarba et al. Journal of Dentistry 143 (2024) 104862

design. Nonetheless, it is crucial to acknowledge persistent limitations in randomized controlled trials, BMC Oral Health 13 (2017) 150, https://doi.org/
10.1186/s12903-017-0441-y.
the application of AI in implant planning software. Currently, there is a
[4] I. García-Gil, J. Cort!es-Bret!
on-Brinkmann, J. Jim!enez-García, J. Pel!aez-Rico, M.-
notable lack of fully automated implant planning software. Further- J. Su!
arez-García, Precision and practical usefulness of intraoral scanners in implant
more, a gap exists in the literature concerning the proper validation of dentistry: A systematic literature review, J. Clin. Exp. Dent. 12 (2020) e784–e793,
the automated steps within the analyzed programs. Ideally, such vali- https://doi.org/10.4317/jced.57025.
[5] T. Joda, M. Ferrari, G.O. Gallucci, J.G. Wittneben, U. Br”agger, Digital technology in
dation should adhere to randomized control protocols, incorporating fixed implant prosthodontics, Periodontol. 73 (2017) 178–192, https://doi.org/
substantial sample sizes and diverse levels of complexity for implant 10.1111/prd.12164.
placement. These limitations underscore the necessity of advancing AI [6] C. Mangano, F. Luongo, M. Migliario, C. Mortellaro, F.G. Mangano, Combining
Intraoral Scans, Cone Beam Computed Tomography and Face Scans: The Virtual
technology through dedicated research projects with the goal of Patient, J. Craniofac. Surg. 29 (2018) 2241–2246, https://doi.org/10.1097/
achieving comprehensive automation in the domain of dental implant SCS.0000000000004485.
planning. [7] R. Jacobs, B. Salmon, M. Codari, B. Hassan, M.M. Bornstein, Cone beam computed
tomography in implant dentistry: recommendations for clinical use, BMC Oral
Moreover, automation-based steps in the studied programs were not Health 18 (2018) 88, https://doi.org/10.1186/s12903-018-0523-5.
cleary defined in terms of their reliance on an AI algorithm. It is essential [8] L. Ritter, S.D. Reiz, D. Rothamel, T. Dreiseidler, V. Karapetian, M. Scheer, J.
to distinguish between automation and AI-based automation, where E. Z”
oller, Registration accuracy of three-dimensional surface and cone beam
computed tomography data for virtual implant planning, Clin. Oral Implants Res.
automation involves predetermined instructions to act in a given situ- 23 (2012) 447–452, https://doi.org/10.1111/j.1600-0501.2011.02159.x.
ation, whereas AI is a computerized system that is designed to recognize [9] C. Mangano, A. Bianchi, F.G. Mangano, J. Dana, M. Colombo, I. Solop, O. Admakin,
patterns and simulate human thinking [78]. Custom-made 3D printed subperiosteal titanium implants for the prosthetic
restoration of the atrophic posterior mandible of elderly patients: a case series, 3D
In terms of future perspectives, novel studies must be carried out in
Print. Med. 6 (2020) 1, https://doi.org/10.1186/s41205-019-0055-x.
conjunction with industry developments to generate substantial evi- [10] C. Mangano, G. Luongo, F. Luongo, H. Lerner, B. Margiani, O. Admakin,
dence supporting the seamless integration of AI into all subsequent F. Mangano, Custom-made computer-aided-design/computer-assisted-
digital procedures, thus yielding a profound impact on clinical practice. manufacturing (CAD/CAM) synthetic bone grafts for alveolar ridge augmentation:
A retrospective clinical study with 3 years of follow-up, J. Dent. 127 (2022)
Furthermore, it is crucial to underscore that the entire implant planning 104323, https://doi.org/10.1016/j.jdent.2022.104323.
process should ideally take place within a single application or platform [11] A. Cucchi, A. Bianchi, P. Calamai, L. Rinaldi, F. Mangano, E. Vignudelli,
to fully harness the advantages AI tools offer. Hence, a follow-up review G. Corinaldesi, Clinical and volumetric outcomes after vertical ridge augmentation
using computer-aided-design/computer-aided manufacturing (CAD/CAM)
can be conducted in the coming five years to explore recent updates in customized titanium meshes: a pilot study, BMC Oral Health 20 (2020) 219,
the literature and advancements in implant planning software/platform, https://doi.org/10.1186/s12903-020-01205-4.
with a focus on the evolving levels of AI-based automation involved. [12] F. Mangano, M. Bazzoli, L. Tettamanti, D. Farronato, M. Maineri, A. Macchi,
C. Mangano, Custom-made, selective laser sintering (SLS) blade implants as a non-
conventional solution for the prosthetic rehabilitation of extremely atrophied
5. Conclusion posterior mandible, Lasers Med. Sci. 28 (2013) 1241–1247, https://doi.org/
10.1007/s10103-012-1205-1.
[13] M. Chung, J. Lee, W. Song, Y. Song, J. Lee, Y.G. Shin, Automatic Registration
The primary use of AI has been identified in various steps of virtual Between Dental Cone-Beam CT and scanned surface via deep pose regression
patient creation, specifically in anatomical landmark segmentation. The neural networks and clustered similarities, IEEE Trans. Med. Imaging 39 (2020)
latter may allow high accuracy, time efficiency, and consistency. Among 3900–3909, https://doi.org/10.1109/TMI.2020.3007520.
[14] S. Shujaat, M. Riaz, R. Jacobs, Synergy between artificial intelligence and precision
the 12 dental implant planning software programs identified, only six
medicine for computer-assisted oral and maxillofacial surgical planning, Clin. Oral
demonstrated partial automation in at least one step of the digital Investig. 27 (2023) 897–906, https://doi.org/10.1007/s00784-022-04706-4.
workflow. A fully automated digital workflow for virtual implant [15] A.F. Leite, K. de F. Vasconcelos, H. Willems, R. Jacobs, Radiomics and machine
learning in oral healthcare, Proteomics Clin. Appl. 14 (2020) e1900040, https://
placement still needs to be developed and scientifically validated. It is
doi.org/10.1002/prca.201900040.
clear that such advancement would facilitate the pre-surgical planning [16] S. Mureșanu, O. Alm# așan, M. Hedeșiu, L. Dioșan, C. Dinu, R. Jacobs, Artificial
phase to be performed by surgeons and dental specialists for time- intelligence models for clinical usage in dentistry with a focus on
efficient and predictable dental implant treatment. dentomaxillofacial CBCT: a systematic review, Oral. Radiol. 39 (2023) 18–40,
https://doi.org/10.1007/s11282-022-00660-9.
[17] B.M. Elgarba, S. Van Aelst, A. Swaity, N. Morgan, S. Shujaat, R. Jacobs, Deep
CRediT authorship contribution statement learning-based segmentation of dental implants on cone-beam computed
tomography images: a validation study, J. Dent. 137 (2023) 104639, https://doi.
org/10.1016/j.jdent.2023.104639.
Bahaaeldeen M. Elgarba: Writing – review & editing, Writing – [18] M. Gerhardt, R.C. Fontenele, A.F. Leite, P. Lahoud, H. Willems, A. Smolders,
original draft, Visualization, Resources, Methodology, Investigation, T. Beznik, R. Jacobs, Automated detection and labelling of teeth and small
Data curation, Conceptualization. Rocharles Cavalcante Fontenele: edentulous regions on cone-beam computed tomography using convolutional
neural networks, J. Dent. 122 (2022) 104139, https://doi.org/10.1016/j.
Writing – review & editing, Visualization, Resources, Methodology. jdent.2022.104139.
Mihai Tarce: Writing – review & editing, Investigation. Reinhilde Ja- [19] F. Preda, N. Morgan, A. Van Gerven, F. Nogueira-Reis, A. Smolders, X. Wang,
cobs: Writing – review & editing, Supervision, Methodology, S. Nomidis, E. Shaheen, H. Willems, R. Jacobs, Deep convolutional neural network-
based automated segmentation of the maxillofacial complex from cone-beam
Conceptualization. computed tomography: A validation study, J. Dent. 124 (2022) 104238, https://
doi.org/10.1016/j.jdent.2022.104238.
Declaration of competing interest [20] P. Verhelst, A. Smolders, T. Beznik, J. Meewis, A. Vandemeulebroucke, E. Shaheen,
A. Van Gerven, H. Willems, C. Politis, R. Jacobs, Layered deep learning for
automatic mandibular segmentation in cone-beam computed tomography, J. Dent.
The authors declare that they have no known competing financial 114 (2021) 103786, https://doi.org/10.1016/j.jdent.2021.103786.
interests or personal relationships that could have appeared to influence [21] P. Lahoud, S. Diels, L. Niclaes, S. Van Aelst, H. Willems, A. Van Gerven,
M. Quirynen, R. Jacobs, Development and validation of a novel artificial
the work reported in this paper. intelligence driven tool for accurate mandibular canal segmentation on CBCT,
J. Dent. 116 (2022) 103891, https://doi.org/10.1016/j.jdent.2021.103891.
References [22] F. Nogueira-Reis, N. Morgan, I.R. Suryani, C.P.M. Tabchoury, R. Jacobs, Full
virtual patient generated by Artificial Intelligence-driven integrated segmentation
of craniomaxillofacial structures from CBCT images, J. Dent. (2023) 104829,
[1] S. Shujaat, M.M. Bornstein, J.B. Price, R. Jacobs, Integration of imaging modalities
https://doi.org/10.1016/j.jdent.2023.104829.
in digital dental workflows - possibilities, limitations, and potential future
[23] A. Saeed, M. Alkhurays, M. AlMutlaqah, M. AlAzbah, SA. Alajlan, Future of using
developments, Dentomaxillofac. Radiol. 50 (2021) 20210268, https://doi.org/
robotic and artificial intelligence in implant dentistry, Cureus 9 (2023) e43209,
10.1259/dmfr.20210268.
https://doi.org/10.7759/cureus.43209.
[2] M. Vercruyssen, I. Laleman, R. Jacobs, M. Quirynen, Computer-supported implant
[24] M. Revilla-Le! on, M. G! omez-Polo, S. Vyas, B.A. Barmak, G.O. Galluci, W. Att, V.
planning and guided surgery: a narrative review, Clin. Oral Implants Res. 26
R. Krishnamurthy, Artificial intelligence applications in implant dentistry: A
(2015) 69–76, https://doi.org/10.1111/clr.12638.
systematic review, J. Prosthet. Dent. 129 (2023) 293–300, https://doi.org/
[3] M. Colombo, C. Mangano, E. Mijiritsky, M. Krebs, U. Hauschild, T. Fortin, Clinical
10.1016/j.prosdent.2021.05.008.
applications and effectiveness of guided implant surgery: a critical review based on

10