Acceptability, deviation, and efficiency of automated artificial intelligence–based implant planning methods: A comparison with human intelligence

Satita Leelaluk; Hanin Hammoudeh; Suchada Kongkiatkamon; Tai‐Hsien Wu; Paola Saponaro; Kedith Sawangsri

doi:10.1111/jopr.70161

. 2026 May 13;35(7):1038–1045. doi: 10.1111/jopr.70161

Acceptability, deviation, and efficiency of automated artificial intelligence–based implant planning methods: A comparison with human intelligence

Satita Leelaluk ¹, Hanin Hammoudeh ^2,³, Suchada Kongkiatkamon ⁴, Tai‐Hsien Wu ⁵, Paola Saponaro ¹, Kedith Sawangsri ^2,^✉

PMCID: PMC13456366 PMID: 42126044

Abstract

Purpose

To evaluate the acceptability, deviation, and efficiency of two automated, artificial intelligence‐driven implant planning methods compared with a human‐based approach for single‐tooth replacement.

Materials and Methods

Data of 32 patients, involving a single edentulous span, were retrospectively obtained, including cone‐beam computed tomography (CBCT) and intraoral scans. Three implant planning methods were applied per case: human‐based method (HP), automated CBCT–image‐based method (AA), and automated CBCT–segmentation‐based method (RL). Implant position acceptability was assessed by three prosthodontists using a standardized three‐point scale on 12 surgical and prosthetic parameters. Using HP as a reference, angular and linear positional deviations were quantified at the implant platform and apex. Total planning time was recorded. Statistical analyses included Friedman's chi‐square test, Wilcoxon test, ANOVA with Tukey's post hoc tests, and intraclass correlation coefficient (α = 0.05).

Results

No significant differences were observed between RL and HP for most surgical and prosthetic parameters except for the crestal bone and platform‐planned gingival margin distance. At the same time, AA showed significantly lower scores than both methods. No significant differences were found among the three methods for implant–mandibular canal distance. AA demonstrated significantly higher positional deviation than RL (p < 0.001). Total planning time differed significantly among methods. HP (14.1 ± 2.44 min), RL (8.91 ± 2.89 min), and AA (3.59 ± 1.44 min) with all pairwise comparisons significant (p < 0.001).

Conclusions

Compared to HP, RL demonstrated similar surgical and prosthetic acceptability with faster planning time. AA was the most efficient but showed lower acceptability than both HP and RL, and demonstrated greater positional deviation than RL.

Keywords: artificial intelligence, dental implant, digital dentistry, implant planning

Dental implants are widely used to rehabilitate partially and completely edentulous patients, and survival rates exceed 90% at 10 years. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ Computer‐aided implant planning enables precise three‐dimensional (3D) positioning relative to anatomical and prosthetic requirements, achieving accuracy rates of 95%–100%. ¹¹ , ¹² , ¹³ , ¹⁴ The workflow includes aligning digital scans with tomographic data, diagnostic waxing, anatomical segmentation, nerve tracing, and virtual implant placement. However, the process remains complex and time‐consuming due to its steep learning curve and variability in patient anatomy and clinician experience. ¹⁰ , ¹¹

Artificial intelligence (AI) has become an integral part of presurgical implant planning. ¹⁵ , ¹⁶ , ¹⁷ Multiple planning platforms have integrated AI to assist tasks such as patient data alignment, diagnostic waxing, thresholding and segmentation of cone‐beam computed tomography (CBCT), and nerve tracing, streamlining the process. ¹⁸ , ¹⁹ , ²⁰ Literature reported that these AI‐assisted applications demonstrated high accuracy exceeding 90%. ¹⁸ , ¹⁹ , ²⁰ Recent advancements have enabled implant planning software to perform the entire workflow, from CBCT thresholding to virtual implant placement, without human input. These fully automated systems can be categorized into two approaches based on their computational designs. The CBCT–image‐based planning approach relies on differential CBCT gray‐value information to infer bone contours, dentition, and anatomical constraints. In addition, the software uses diagnostic scans and AI‐driven virtual waxing to provide prosthetic information for virtual implant positioning. In contrast, the CBCT–segmentation‐based planning approach first generates 3D reconstructions of bone, teeth, gingiva, and anatomical structures using CBCT thresholding and segmentation algorithms. Then, the program performs virtual waxing and implant positioning within 3D‐segmented anatomical boundaries and the planned prosthetic space. Previous studies using experimental CBCT–segmentation‐based implant planning have demonstrated expert‐level accuracy for single posterior mandibular implant planning. ²¹ , ²² However, evidence regarding the acceptability, accuracy, and efficiency of commercial AI‐driven virtual implant planning platforms remains limited. The present retrospective study aimed to assess the acceptability, deviation, and efficiency of two automated AI‐based implant planning methods for a single edentulous span compared to a human‐based method (HP). The null hypothesis was that no differences would be observed in acceptability, deviation, or efficiency between the two AI‐based methods and HPs.

MATERIALS AND METHODS

This retrospective study was approved by the Institutional Review Board (IRB No. 2024E1228). A dataset of 32 CBCT scans and corresponding maxillary and mandibular intraoral scans (IOSs) from patients treated at a university dental clinic for single edentulous span implant therapy was collected.

CBCT scans were acquired using the Planmeca Viso G7 system (Planmeca) with a 200‐µm voxel size and a Ø110 × 110‐mm field of view; the CBCT scans were obtained at 90 kVp and 10 mA with a 9‐s exposure time, using the standard dose mode and an implant reconstruction filter. Intraoral scans were obtained using the Trios 4 scanner (3Shape A/S). Eligible cases met the following inclusion criteria: single edentulous span adjacent to at least one natural tooth; scans taken more than 3 months post‐extraction; healthy bone dimensions of at least 12 mm height and 7 mm width. Exclusion criteria included scans with motion artifacts; severe beam‐hardening artifact preventing IOS–CBCT registration; incomplete anatomical capture, such as the absence of the mandibular canal, maxillary sinus, or nasal floor; incomplete arch capture; insufficient bone quality; proximity to existing implants; recent extraction (less than 3 months); bone augmentation performed within the preceding 3 months; or the presence of pathology.

CBCT scans were exported as Digital Imaging and Communications in Medicine (DICOM) files, and intraoral scans as Standard Tessellation Language (STL) files. Three implant planning methods were evaluated (A graphical abstract is provided in Figure 1):

Human‐based method (HP): A single experienced operator (P.S., with 10 years of experience in implant dentistry) performed diagnostic waxing on the articulated IOSs in dental CAD software (Dental System 2023, 3Shape A/S) and exported the IOS with diagnostic waxing as an STL file. DICOM data and IOSs with diagnostic waxing were imported into the desktop‐based implant planning software (RealGUIDE Z3D version 5.4.2, Zimvie). The same operator performed preparatory steps, including CBCT segmentation via thresholding, CBCT–IOS registration using iterative closest point (ICP) best‐fit algorithms, mandibular canal tracing, and implant selection (Straumann BLT SLA, Straumann). The implants were positioned according to a non‐angled, screw‐retained prosthetic design, biologic requirements, and available bone. The planned implant position and IOS were exported together for deviation analysis (Figure 2a).
Automated CBCT–image‐based method (AA): The operator (S.L.) uploaded the DICOM file and IOSs to the desktop‐based implant planning software (Atomica Implant Planner version 8.2.0, Atomica Technology). AI‐driven automation performed thresholding segmentation, CBCT–IOS registration, and mandibular canal tracing without operator intervention. The platform required the operator to identify the edentulous site and select the implant system and dimensions. The same implant system and dimensions were used in the HP group to ensure standardization. The software automatically generated diagnostic waxing and implant placement based on uploaded intraoral and CBCT images. The planned implant position and IOS were exported together for deviation analysis (Figure 2b).
Automated CBCT–segmentation‐based method (RL): The operator (S.L.) uploaded the DICOM file and IOS to the cloud‐based implant planning platform (Relu Creator version 5.0.2, Relu). The platform used AI‐driven automation to perform thresholding segmentation and 3D reconstruction of dentition, bone, gingiva, inferior alveolar nerve, and maxillary sinuses; CBCT–IOS registration; and mandibular canal tracing without operator intervention, based on the AI model's training, the reconstructed model, and anatomical requirements. ²² For standardization, implant dimensions were manually modified to match the HP group before being exported along with the IOS for deviation analysis (Figure 2c).

Graphical abstract summarizing the study workflow. Cone‐beam computed tomography (CBCT) scans and intraoral scans (IOSs) were acquired and used as input data across three implant planning methods: human‐based (HP), automated CBCT–image‐based (AA), and automated CBCT–segmentation‐based (RL). The implant position output of each method was evaluated based on acceptability, positional deviation, and efficiency.

Implant planning methods were evaluated in this study. (a) Human‐based planning method (HP); (b) automated CBCT–image‐based planning method (AA); (c) automated CBCT–segmentation‐based planning method (RL).

For acceptability assessment, the IOSs with implant position from three approaches were uploaded to the desktop‐based implant planning software (RealGUIDE Z3D 5.4.2, Zimvie) and superimposed using the ICP best‐fit surface of the IOSs (Figure 3). Three calibrated prosthodontists (K.S., H.H., and S.K.), masked to the planning approaches, independently rated each implant positioning based on surgical criteria (implant relationship to crestal bone, distance to buccal/lingual bone, adjacent tooth, sinus floor, and mandibular canal) and prosthetic criteria (platform relationship to planned gingival margin, bucco‐lingual and mesio‐distal position, and angulation in relation to planned prosthesis). Ratings followed a standardized three‐point scale (2 = ideal; 1 = satisfactory/minor adjustment required; 0 = unacceptable/major adjustment required) according to established evaluation guidelines (Supporting Information S1). Before evaluation, the three prosthodontist graders participated in a calibration session reviewing the scoring criteria and evaluation parameters. The calibration procedure is described in Supporting Information S2.

(From left to right) Representative radiographs of cross‐sectional, axial, and tangential views comparing implant positions planned by three methods: human‐based (HP, red), automated CBCT–image‐based (AA, green), and automated CBCT–segmentation‐based (RL, purple), in two anatomical regions: (a) maxillary anterior and (b) mandibular posterior.

A custom computational workflow was developed in a Python environment using the vedo library, which is built on the Visualization Toolkit (VTK), to perform all deviation analyses. The IOS datasets containing the planned implant positions were imported, and an ICP best‐fit alignment was executed to register each experimental dataset (AA and/or RL) to the reference dataset (HP). ICP iteratively refines the transformation matrix to minimize point‐to‐point discrepancies between meshes, enabling reproducible superimposition without manual intervention. To standardize measurement points, the mesh topology of each implant file was examined, and the vertex indices corresponding to the platform center and apical tip were determined in advance for each software export. The program then extracted these coordinates directly from the registered models, allowing automated computation of all deviation outcomes. These included 3D angular deviation (degrees) between the implant long axes in full spatial orientation; 3D linear deviation (mm) at the platform and apex; lateral deviation (mm) at both levels; and vertical deviation (mm) along the implant's long axis (Figure 4). ²³ All calculations were fully automated to reduce operator‐dependent variability. Accuracy of the measurements was internally validated against commercial 3D inspection software (Geomagic Control X; 3D Systems).

Schematic illustration of implant positional deviation between the reference (HP) and experimental (AA or RL) planning methods: (1) angular deviation between implant long axes (degrees); (2.1 and 2.2) three‐dimensional linear deviation at platform and apex levels (mm); (3.1 and 3.2) lateral linear deviation at platform and apex levels (mm); and (4.1 and 4.2) vertical deviation at platform and apex levels (mm).

For efficiency assessment, planning time for each approach was recorded in minutes, starting from the treatment setup and data upload through diagnostic waxing procedure, DICOM segmentation, IOS–CBCT registration, mandibular canal tracing, and implant selection/positioning. Time required for saving, exporting, and transferring STL files was excluded. All implant planning was conducted on a standardized computer system equipped with an Intel Core i9‐11,900 CPU, an NVIDIA GeForce RTX 3050 Ti Laptop GPU with 4GB of GDDR6 memory, and 32GB of RAM.

Acceptability scores were analyzed using Friedman's chi‐square test to compare median acceptability ranks among the three implant planning methods for each evaluated aspect. To control multiple comparisons across aspects, Bonferroni‐adjusted p values were calculated. A post hoc sensitivity analysis was performed (G*Power software version 3.1.9.7; Heinrich Heine University). Because the primary analysis used Friedman's chi‐square test for ordinal outcomes across three planning methods, design sensitivity was estimated using a generic chi‐square model with 2 degrees of freedom. The significance level was adjusted to α = 0.05/12 to account for multiple evaluated parameters. Intraclass correlation coefficient (ICC) was calculated. Normality and homogeneity of variances of implant deviation and planning time were assessed using the Shapiro–Wilk and Levene's tests, respectively. Implant deviation and planning time across all platforms were analyzed using the Wilcoxon signed‐rank test and one‐way ANOVA with Tukey's post hoc test, respectively. All statistical analyses were conducted using a statistical software program (SAS software, Version 9.4 SAS Institute Inc.) with an overall significance level of α = 0.05

RESULTS

After analyzing intraoral and tomographic data from 32 single edentulous spans that met the inclusion criteria, 16 were in the anterior region (13 maxillary, 3 mandibular) and 16 were in the posterior region (6 maxillary, 10 mandibular). Under these assumptions, the available sample size provided >95% statistical power to detect a large effect size (w ≥ 0.80) for all parameters except the implant–mandibular canal distance.

Inter‐rater reliability among the three evaluators was excellent (ICC(2,3) = 0.99; 95% CI: 0.983–0.993). Using Friedman's chi‐square test on ordinal acceptability scores, HP scored significantly higher than RL for implant position relative to the crestal bone and platform‐planned gingival margin distance (p < 0.05). Compared with AA, HP demonstrated significantly higher scores for all evaluated parameters except implant–mandibular canal distance (p < 0.05). No significant differences were observed between HP and RL for most surgical and prosthetic parameters, including implant–buccal bone distance, implant–tooth distance (mesial and distal), sinus/nasal floor distance, bucco–lingual position, mesio–distal position, and bucco–lingual or mesio–distal angulation. AA showed significantly lower acceptability scores than both HP and RL for implant position relative to the crestal bone, implant–buccal bone distance, implant–tooth distance (mesial), platform–planned gingival margin distance, bucco–lingual position, and bucco–lingual and mesio–distal angulation (p < 0.05). No significant differences were found among the three methods for implant–mandibular canal distance (p > 0.05). Descriptive statistics of the acceptability scores of implant planning methods among the three groups are presented in Table 1

TABLE 1.

Acceptability score of implant position by groups shown in median (interquartile range).

	Parameter	Abbreviation	HP	AA	RL
Surgical parameter	Implant position relative to crestal bone	C bone	2 (2–2)^a	0 (0–1)^c	2 (1–2)^b
	Implant–buccal bone distance	B bone	2 (2–2)^a	1 (1–2)^b	2 (1–2)^a
	Implant–lingual bone distance	L bone	2 (1–2)^a	1 (1–1)^b	1 (1–2)^a ^b
	Implant–tooth distance (mesial)	I‐T mesial	2 (2–2)^a	1 (1–2)^b	2 (1–2)^a
	Implant–tooth distance (distal)	I‐T distal	2 (2–2)^a	1 (1–2)^b	2 (1–2)^a ^b
	Implant–sinus/nasal floor distance	Sinus floor	2 (2–2)^a	1 (1–2)^b	2 (1–2)^a ^b
	Implant–mandibular canal distance	Mn canal	2 (1–2)	2 (2–2)	2 (2–2)
Prosthetic parameters	Platform–planned gingival margin distance	GM distance	2 (2–2)^a	0 (0–1)^c	1 (1–2)^b
	Bucco–lingual position	BL position	2 (2–2)^a	1 (1–1)^b	2 (1–2)^a
	Mesio–distal position	MD position	2 (2–2)^a	1 (1–2)^b	2 (1–2)^a ^b
	Bucco–lingual angulation	BL angulation	2 (2–2)^a	1 (1–1)^b	2 (1–2)^a
	Mesio–distal angulation	MD angulation	2 (2–2)^a	1 (1–2)^b	2 (1–2)^a

Open in a new tab

Note: Within each row, values indicated by different superscript letters are significantly different at α = 0.05, in descending score order (a > b > c).

Abbreviations: AA, automated CBCT–image‐based method; CBCT, cone‐beam computed tomography; HP, human‐based method; RL, automated CBCT–segmentation‐based method.

Descriptive statistics for positional deviation are shown in Table 2 and Supporting Information S3. Wilcoxon signed‐rank tests indicated that RL produced significantly smaller deviations than AA across all outcomes, including angular deviation, 3D deviation, lateral deviation, and vertical deviation for both platform and apex levels (p < 0.001).

TABLE 2.

Angular and linear deviation of implant position are shown in mean ± standard deviation.

Implant deviation	AA	RL	p values
Angular deviation (°)	20.48 ± 21.81	5.60 ± 3.72	<0.001*
Platform level deviation (mm)
3D deviation	3.18 ± 1.88	1.33 ± 0.84	<0.001*
Vertical deviation	2.50 ± 1.34	0.79 ± 0.82	<0.001*
Lateral deviation	1.76 ± 1.59	0.86 ± 0.66	<0.001*
Apex level deviation (mm)
3D deviation	5.71 ± 3.53	1.80 ± 1.18	<0.001*
Lateral deviation	3.47 ± 1.94	0.52 ± 1.34	<0.001*
Vertical deviation	4.27 ± 3.32	1.22 ± 1.06	<0.001*

Open in a new tab

Abbreviations: 3D, three‐dimensional; AA, automated CBCT–image‐based method; CBCT, cone‐beam computed tomography; RL, automated CBCT–segmentation‐based method.

Total planning time, illustrated in Table 3 and Supporting Information S4, differed significantly among the three programs (one‐way ANOVA), F(2,93) = 161, p < 0.001. HP required the longest time (14.1 ± 2.44 min), RL was intermediate (8.91 ± 2.89 min), and AA was the fastest (3.59 ± 1.44 min). Tukey's test showed all pairwise differences were significant, where HP–AA mean differencewas 10.5 min (p < 0.001), HP–RL mean differencewas 5.17 min (p < 0.001), and RL–AA mean differencewas 5.32 min (p < 0.001).

TABLE 3.

Planning time by groups (in minutes) shown in mean ± standard deviation.

Group	Diagnostic waxing	IOS‐CBCT preparation	Implant positioning	Total planning time
HP	4.23 ± 1.13	6.88 ± 1.24	2.99 ± 1.07	14.1 ± 2.44^a
AA	‐	2.91 ± 0.62	0.68 ± 0.42	3.59 ± 1.44^b
RL	‐	‐	‐	8.91 ± 2.89^c

Open in a new tab

Note: Values indicated by the same superscript in each column letter were not significantly different at α = 0.05.

Abbreviations: AA, automated CBCT–image‐based method; CBCT, cone‐beam computed tomography; HP, human‐based method; IOS, intraoral scan; RL, automated CBCT–segmentation‐based method.

DISCUSSION

The results of this study revealed significant differences in acceptability, deviation, or efficiency between the AI‐based and the HP methods. Therefore, the null hypothesis that no differences would be observed in acceptability, deviation, or efficiency between the two AI‐based and the HP methods was rejected.

This study demonstrated that the implant planning method significantly influences acceptability. These results are consistent with prior AI–human comparisons which showed that the RL method was largely comparable to HP in overall acceptability, ²² whereas AA underperformed in several domains. Especially, RL's acceptability matched HP for most surgical and prosthetic parameters, except for implant position relative to the crestal bone and implant platform–planned gingival margin distance, while HP showed significantly higher scores than both AI‐based methods. Prior studies show implant positioning may vary based on planner training and specialty, ²⁴ , ²⁵ which may contextualize the results; in this study, the planners and evaluators were prosthodontists, and the planned restorations were non‐angled, screw‐retained, potentially favoring HP on prosthetic parameters. Moreover, the observed differences between AA and RL are likely due to different workflow designs: RL relies on explicit 3D segmentation and model‐based constraints, which enforce hard anatomic constraints during automated implant placement; ²⁶ AA operates primarily on CBCT grey‐value information without structured segmentation, making implant positioning more susceptible to variability.

This study also demonstrated that the implant positional deviation was significantly influenced by the implant planning method. A previous study evaluating AI versus human‐based planning in the posterior mandible reported mean 3D platform deviations of 1.39 ± 0.65 mm and apical deviations of 1.70 ± 0.79 mm. These results are consistent with the deviations observed in the present study's RL group, despite broader anatomical inclusion criteria. ²² Additionally, these results align with prior findings on inter‐operator variability in virtual implant planning, which reported deviations of 1.11 ± 0.55 mm at the platform, 1.54 ± 0.66 mm at the apex, and an angular deviation of 3.3° ± 1.99°. ²⁴ These comparisons suggest that the RL remains within clinically acceptable thresholds. In contrast, the AA exhibited deviations that exceeded commonly accepted limits. The increased deviation from HP associated with the image‐based AI planning method may be attributed to its reliance on CBCT gray‐value thresholding without structured 3D segmentation or anatomical referencing. While this approach offers greater speed, the lack of precise anatomical constraints increases susceptibility to errors from image artifacts and misinterpretation of structural morphology, resulting in greater angular and vertical deviations compared to segmentation‐based workflows.

Total planning time is a critical measure of clinical efficiency. In this study, the HP method required the longest time (14.1 ± 2.4 min), whereas RL reduced planning time by about 40% (8.9 ± 2.9 min) and AA by nearly 75% (3.6 ± 1.4 min). These findings align with previous reports where AI‐driven implant planning completed plans in 3.2 min compared to 6.8 min for humans, effectively halving planning time. ²² The results also showed that in HP methods, preparatory steps consumed 70%–80% of total planning time, while manual implant positioning itself required approximately 3 min. Since both AI methods still require minor manual adjustment, the efficiency of the AA approach may remain clinically acceptable despite its increased deviation from HP, particularly for straightforward cases where speed and workflow efficiency are prioritized.

Beyond comparing performance among planning approaches, this study highlights the potential complementarity between automated and human‐driven workflows. AI‐based methods can efficiently automate tasks such as anatomical segmentation, edentulous‐site identification, and initial implant positioning, yet they are not sufficiently reliable for independent clinical decision‐making in prosthetically demanding or anatomically complex cases. Currently, these systems serve best as an adjunct, expediting virtual patient set‐up and providing a consistent foundation for clinician refinement. Such hybrid workflows improve efficiency without compromising the clinical judgment required for high‐quality implant planning.

This study has several limitations. It was a virtual planning study using existing CBCT and IOS data with a small sample size and was limited to single edentulous spans, reducing its generalizability across clinical scenarios. The influence of implant location (e.g., maxillary anterior vs. posterior, mandibular anterior vs. posterior) or implant size, length, and geometry was not analyzed, despite their potential impact on planning accuracy and anatomical constraints. Only one planning software and two AI algorithms were evaluated, and the outcomes focused solely on virtual planning, without assessment of surgical transfer accuracy, prosthetic fit, or long‐term outcomes. Although acceptability scoring was performed by calibrated prosthodontists, subjectivity may persist. Additionally, user experience, learning curve, and cost implications were not examined. Future studies should incorporate diverse implant geometries, edentulous sites, larger clinical datasets, and prospective evaluation of surgical, prosthetic, and economic outcomes across multiple AI platforms.

CONCLUSIONS

RL showed no significant differences in acceptability from HP for most surgical and prosthetic parameters. RL demonstrated significantly higher acceptability and lower positional deviation than the AA, which consistently scored lower than both HP and RL. Both AA and RL required significantly less planning time than HP. Although AA was the fastest, RL might provide a more favorable trade‐off between time efficiency and positional deviation.

CONFLICT OF INTEREST STATEMENT

The authors report no conflicts of interest related to this study.

Supporting information

Supplementary Material 1. Criteria for scoring implant positional acceptability based on surgical and prosthetic parameters

JOPR-35-1038-s001.docx^{(17.1KB, docx)}

jopr70161‐sup‐0002‐SuppMat.docx

JOPR-35-1038-s003.docx^{(15.8KB, docx)}

jopr70161‐sup‐0003‐SuppMat.docx

JOPR-35-1038-s004.docx^{(213.3KB, docx)}

jopr70161‐sup‐0004‐SuppMat.docx

JOPR-35-1038-s002.docx^{(66.8KB, docx)}

ACKNOWLEDGMENTS

The authors thank Dr. William M. Johnston, PhD, for his contributions to statistical consultation and data analysis.

Leelaluk S, Hammoudeh H, Kongkiatkamon S, Wu T‐H, Saponaro P, Sawangsri K. Acceptability, deviation, and efficiency of automated artificial intelligence–based implant planning methods: A comparison with human intelligence. J Prosthodont. 2026;35:1038–1045. 10.1111/jopr.70161

This research was previously presented at the John J. Sharry Prosthodontic Research Competition during the American College of Prosthodontists 55th Annual Session, held in New Orleans, Louisiana, USA, on October 23, 2025.

REFERENCES

1. Jemt T, Johansson J. Implant treatment in the edentulous maxillae: A 15‐year follow‐up study on 76 consecutive patients provided with fixed prostheses. Clin Implant Dent Relat Res. 2006;8:61–69. [DOI] [PubMed] [Google Scholar]
2. Krebs M, Schmenger K, Neumann K, Weigl P, Moser W, Nentwig G‐H. Long‐term evaluation of ANKYLOS® dental implants, part I: 20‐year life table analysis of a longitudinal study of more than 12,500 implants. Clin Implant Dent Relat Res. 2015;17:e275–e286. [DOI] [PubMed] [Google Scholar]
3. Kwon T, Bain PA, Levin L. Systematic review of short‐ (5–10 years) and long‐term (10 years or more) survival and success of full‐arch fixed dental hybrid prostheses and supporting implants. J Dent. 2014;42:1228–1241. [DOI] [PubMed] [Google Scholar]
4. Pjetursson BE, Brägger U, Lang NP, Zwahlen M, Comparison of survival and complication rates of tooth‐supported fixed dental prostheses (FDPs) and implant‐supported FDPs and single crowns (SCs) [published correction appears in]. Clin Oral Implants Res. 2007;18:97–113. [DOI] [PubMed] [Google Scholar]
5. Sailer I, Barbato L, Mojon P, Pagliaro U, Serni L, Karasan D, et al. Single and partial tooth replacement with fixed dental prostheses supported by dental implants: A systematic review of outcomes and outcome measures used in clinical trials in the last 10 years. J Clin Periodontol. 2023;34:22–37. [DOI] [PubMed] [Google Scholar]
6. Ferrigno N, Laureti M, Fanali S, Grippaudo G. A long‐term follow‐up study of non‐submerged ITI implants in the treatment of totally edentulous jaws. Part I: Ten‐year life table analysis of a prospective multicenter study with 1286 implants. Clin Oral Implants Res. 2002;13:260–273. [DOI] [PubMed] [Google Scholar]
7. Gallucci GO, Doughtie CB, Hwang JW, Fiorellini JP, Weber H‐P. Five‐year results of fixed implant‐supported rehabilitations with distal cantilevers for the edentulous mandible. Clin Oral Implants Res. 2009;20:601–607. [DOI] [PubMed] [Google Scholar]
8. Jemt T, Häger P. Early complete failures of fixed implant‐supported prostheses in the edentulous maxilla: A 3‐year analysis of 17 consecutive cluster failure patients. Clin Implant Dent Relat Res. 2006;8:77–86. [DOI] [PubMed] [Google Scholar]
9. Örtorp A, Jemt T. Early laser‐welded titanium frameworks supported by implants in the edentulous mandible: A 15‐year comparative follow‐up study. Clin Implant Dent Relat Res. 2009;11:311–322. [DOI] [PubMed] [Google Scholar]
10. Patzelt SBM, Bahat O, Reynolds MA, Strub JR. The all‐on‐four treatment concept: A systematic review. Clin Implant Dent Relat Res. 2014;16:836–855. [DOI] [PubMed] [Google Scholar]
11. Pellegrino G, Ferri A, Del Fabbro M, Prati C, Gandolfi M, Marchetti C. Dynamic navigation in implant dentistry: A systematic review and meta‐analysis. Int J Oral Maxillofac Implants. 2021;36:e121–e140. [DOI] [PubMed] [Google Scholar]
12. Tahmaseb A, Wu V, Wismeijer D, Coucke W, Evans C. The accuracy of static computer‐aided implant surgery: A systematic review and meta‐analysis. Clin Oral Implants Res. 2018;29:416–435. [DOI] [PubMed] [Google Scholar]
13. Colombo M, Mangano C, Mijiritsky E, Krebs M, Hauschild U, Fortin T. Clinical applications and effectiveness of guided implant surgery: A critical review based on randomized controlled trials. BMC Oral Health. 2017;13:150. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Dolcini GA, Colombo M, Mangano C. From guided surgery to final prosthesis with a fully digital procedure: A prospective clinical study on 15 partially edentulous patients. Int J Dent. 2016;2016:7358423. [DOI] [PMC free article] [PubMed] [Google Scholar]
15. Elgarba BM, Fontenele RC, Tarce M, Jacobs R, Artificial intelligence serving pre‐surgical digital implant planning: A scoping review. J Dent. 2024;143:104862. [DOI] [PubMed] [Google Scholar]
16. Gerhardt MDN, Fontenele RC, Leite AF, Lahoud P, Van Gerven A, Willems H, et al. Automated detection and labelling of teeth and small edentulous regions on cone‐beam computed tomography using convolutional neural networks. J Dent. 2022;122:104139. [DOI] [PubMed] [Google Scholar]
17. Verhelst P‐J, Smolders A, Beznik T, Meewis J, Vandemeulebroucke A, Shaheen E, et al. Layered deep learning for automatic mandibular segmentation in cone‐beam computed tomography. J Dent. 2021;114:103786. [DOI] [PubMed] [Google Scholar]
18. Jaskari J, Sahlsten J, Järnstedt J, Mehtonen H, Karhu K, Sundqvist O, et al. Deep learning method for mandibular canal segmentation in dental cone beam computed tomography volumes. Sci Rep. 2020;10:5842–5842. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. Polizzi A, Quinzi V, Ronsivalle V, Venezia P, Santonocito S, Lo Giudice A, et al. Tooth automatic segmentation from CBCT images: A systematic review. Clin Oral Investig. 2023;27:3363–3378. [DOI] [PubMed] [Google Scholar]
20. Revilla‐León M, Gohil A, Barmak AB, Zandinejad A, Raigrodski AJ, Alonso Pérez‐Barquero J. Best‐fit algorithm influences on virtual casts' alignment discrepancies. J Prosthodont. 2023;32:331–339. [DOI] [PubMed] [Google Scholar]
21. Elgarba BM, Fontenele RC, Ali S, Swaity A, Meeus J, Shujaat S, et al. Validation of a novel AI‐based automated multimodal image registration of CBCT and intraoral scan aiding presurgical implant planning. Clin Oral Implants Res. 2024;35:1506–1517. [DOI] [PubMed] [Google Scholar]
22. Elgarba BM, Fontenele RC, Du X, Mureșanu S, Tarce M, Meeus J, et al. Artificial intelligence versus human intelligence in presurgical implant planning: A preclinical validation. Clin Oral Implants Res. 2025;36:835–845. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Zhao W, Teng W, Su Y, Zhou L. Accuracy of dental implant surgery with freehand, static computer‐aided, dynamic computer‐aided, and robotic computer‐aided implant systems: An in vitro study. J Prosthet Dent. 2024. [DOI] [PubMed] [Google Scholar]
24. Dragan IF, Pirc M, Rizea C, Yao J, Acharya A, Mattheos N, et al. A global perspective on implant education: Cluster analysis of the ‘first dental implant experience’ of dentists from 84 nationalities. Eur J Dent Educ. 2019;23:251–265. [DOI] [PubMed] [Google Scholar]
25. Jayachandran S, Walmsley AD, Hill K: Challenges in dental implant provision and its management in general dental practice. J Dent. 2020;99:103414. [DOI] [PubMed] [Google Scholar]
26. Elgarba BM, Fontenele RC, Mangano F, Jacobs R. Novel AI‐based automated virtual implant placement: Artificial versus human intelligence. J Dent. 2024;147:105146. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1. Criteria for scoring implant positional acceptability based on surgical and prosthetic parameters

JOPR-35-1038-s001.docx^{(17.1KB, docx)}

jopr70161‐sup‐0002‐SuppMat.docx

JOPR-35-1038-s003.docx^{(15.8KB, docx)}

jopr70161‐sup‐0003‐SuppMat.docx

JOPR-35-1038-s004.docx^{(213.3KB, docx)}

jopr70161‐sup‐0004‐SuppMat.docx

JOPR-35-1038-s002.docx^{(66.8KB, docx)}

[jopr70161-bib-0001] 1. Jemt T, Johansson J. Implant treatment in the edentulous maxillae: A 15‐year follow‐up study on 76 consecutive patients provided with fixed prostheses. Clin Implant Dent Relat Res. 2006;8:61–69. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0002] 2. Krebs M, Schmenger K, Neumann K, Weigl P, Moser W, Nentwig G‐H. Long‐term evaluation of ANKYLOS® dental implants, part I: 20‐year life table analysis of a longitudinal study of more than 12,500 implants. Clin Implant Dent Relat Res. 2015;17:e275–e286. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0003] 3. Kwon T, Bain PA, Levin L. Systematic review of short‐ (5–10 years) and long‐term (10 years or more) survival and success of full‐arch fixed dental hybrid prostheses and supporting implants. J Dent. 2014;42:1228–1241. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0004] 4. Pjetursson BE, Brägger U, Lang NP, Zwahlen M, Comparison of survival and complication rates of tooth‐supported fixed dental prostheses (FDPs) and implant‐supported FDPs and single crowns (SCs) [published correction appears in]. Clin Oral Implants Res. 2007;18:97–113. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0005] 5. Sailer I, Barbato L, Mojon P, Pagliaro U, Serni L, Karasan D, et al. Single and partial tooth replacement with fixed dental prostheses supported by dental implants: A systematic review of outcomes and outcome measures used in clinical trials in the last 10 years. J Clin Periodontol. 2023;34:22–37. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0006] 6. Ferrigno N, Laureti M, Fanali S, Grippaudo G. A long‐term follow‐up study of non‐submerged ITI implants in the treatment of totally edentulous jaws. Part I: Ten‐year life table analysis of a prospective multicenter study with 1286 implants. Clin Oral Implants Res. 2002;13:260–273. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0007] 7. Gallucci GO, Doughtie CB, Hwang JW, Fiorellini JP, Weber H‐P. Five‐year results of fixed implant‐supported rehabilitations with distal cantilevers for the edentulous mandible. Clin Oral Implants Res. 2009;20:601–607. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0008] 8. Jemt T, Häger P. Early complete failures of fixed implant‐supported prostheses in the edentulous maxilla: A 3‐year analysis of 17 consecutive cluster failure patients. Clin Implant Dent Relat Res. 2006;8:77–86. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0009] 9. Örtorp A, Jemt T. Early laser‐welded titanium frameworks supported by implants in the edentulous mandible: A 15‐year comparative follow‐up study. Clin Implant Dent Relat Res. 2009;11:311–322. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0010] 10. Patzelt SBM, Bahat O, Reynolds MA, Strub JR. The all‐on‐four treatment concept: A systematic review. Clin Implant Dent Relat Res. 2014;16:836–855. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0011] 11. Pellegrino G, Ferri A, Del Fabbro M, Prati C, Gandolfi M, Marchetti C. Dynamic navigation in implant dentistry: A systematic review and meta‐analysis. Int J Oral Maxillofac Implants. 2021;36:e121–e140. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0012] 12. Tahmaseb A, Wu V, Wismeijer D, Coucke W, Evans C. The accuracy of static computer‐aided implant surgery: A systematic review and meta‐analysis. Clin Oral Implants Res. 2018;29:416–435. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0013] 13. Colombo M, Mangano C, Mijiritsky E, Krebs M, Hauschild U, Fortin T. Clinical applications and effectiveness of guided implant surgery: A critical review based on randomized controlled trials. BMC Oral Health. 2017;13:150. [DOI] [PMC free article] [PubMed] [Google Scholar]

[jopr70161-bib-0014] 14. Dolcini GA, Colombo M, Mangano C. From guided surgery to final prosthesis with a fully digital procedure: A prospective clinical study on 15 partially edentulous patients. Int J Dent. 2016;2016:7358423. [DOI] [PMC free article] [PubMed] [Google Scholar]

[jopr70161-bib-0015] 15. Elgarba BM, Fontenele RC, Tarce M, Jacobs R, Artificial intelligence serving pre‐surgical digital implant planning: A scoping review. J Dent. 2024;143:104862. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0016] 16. Gerhardt MDN, Fontenele RC, Leite AF, Lahoud P, Van Gerven A, Willems H, et al. Automated detection and labelling of teeth and small edentulous regions on cone‐beam computed tomography using convolutional neural networks. J Dent. 2022;122:104139. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0017] 17. Verhelst P‐J, Smolders A, Beznik T, Meewis J, Vandemeulebroucke A, Shaheen E, et al. Layered deep learning for automatic mandibular segmentation in cone‐beam computed tomography. J Dent. 2021;114:103786. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0018] 18. Jaskari J, Sahlsten J, Järnstedt J, Mehtonen H, Karhu K, Sundqvist O, et al. Deep learning method for mandibular canal segmentation in dental cone beam computed tomography volumes. Sci Rep. 2020;10:5842–5842. [DOI] [PMC free article] [PubMed] [Google Scholar]

[jopr70161-bib-0019] 19. Polizzi A, Quinzi V, Ronsivalle V, Venezia P, Santonocito S, Lo Giudice A, et al. Tooth automatic segmentation from CBCT images: A systematic review. Clin Oral Investig. 2023;27:3363–3378. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0020] 20. Revilla‐León M, Gohil A, Barmak AB, Zandinejad A, Raigrodski AJ, Alonso Pérez‐Barquero J. Best‐fit algorithm influences on virtual casts' alignment discrepancies. J Prosthodont. 2023;32:331–339. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0021] 21. Elgarba BM, Fontenele RC, Ali S, Swaity A, Meeus J, Shujaat S, et al. Validation of a novel AI‐based automated multimodal image registration of CBCT and intraoral scan aiding presurgical implant planning. Clin Oral Implants Res. 2024;35:1506–1517. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0022] 22. Elgarba BM, Fontenele RC, Du X, Mureșanu S, Tarce M, Meeus J, et al. Artificial intelligence versus human intelligence in presurgical implant planning: A preclinical validation. Clin Oral Implants Res. 2025;36:835–845. [DOI] [PMC free article] [PubMed] [Google Scholar]

[jopr70161-bib-0023] 23. Zhao W, Teng W, Su Y, Zhou L. Accuracy of dental implant surgery with freehand, static computer‐aided, dynamic computer‐aided, and robotic computer‐aided implant systems: An in vitro study. J Prosthet Dent. 2024. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0024] 24. Dragan IF, Pirc M, Rizea C, Yao J, Acharya A, Mattheos N, et al. A global perspective on implant education: Cluster analysis of the ‘first dental implant experience’ of dentists from 84 nationalities. Eur J Dent Educ. 2019;23:251–265. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0025] 25. Jayachandran S, Walmsley AD, Hill K: Challenges in dental implant provision and its management in general dental practice. J Dent. 2020;99:103414. [DOI] [PubMed] [Google Scholar]

[jopr70161-bib-0026] 26. Elgarba BM, Fontenele RC, Mangano F, Jacobs R. Novel AI‐based automated virtual implant placement: Artificial versus human intelligence. J Dent. 2024;147:105146. [DOI] [PubMed] [Google Scholar]

PERMALINK

Acceptability, deviation, and efficiency of automated artificial intelligence–based implant planning methods: A comparison with human intelligence

Satita Leelaluk, DDS

Hanin Hammoudeh, DDS, MS, FACP

Suchada Kongkiatkamon, DDS, MS, PhD

Tai‐Hsien Wu, PhD

Paola Saponaro, DDS, MS, FACP

Kedith Sawangsri, DDS, MS