Accuracy of robot-assisted dental implant surgery and its clinical influencing factors: a retrospective case series of 100 patients

Abstract

Background

This retrospective study aimed to investigate the accuracy of robot-assisted implant surgery and identify the factors influencing it.

Methods

Patients with single or multiple missing teeth were enrolled in the robot-assisted implant surgery. The patients underwent cone-beam computed tomography (CBCT) using a marker. Virtual implant placement and a drilling sequence were planned prior to surgery. The robotic arm automatically performed the implant osteotomy and placement under the surgeon’s supervision. A postoperative CBCT scan was performed to evaluate deviations between the planned and placed implants.

Results

A total of 152 implants were successfully inserted into the jawbones of 100 patients without any adverse surgical events. However, two implants were observed to fail early (< 1 month). An overall coronal deviation of 0.51 ± 0.02 mm, apical deviation of 0.53 ± 0.02 mm, and angular deviation of 1.05 ± 0.05° were observed. The angular deviation of the first nearest site in the free-end position was significantly lower than that of the second or third nearest site (p = 0.043). Additionally, the coronal and apical deviations were significantly decreased with the accumulation of clinical experience (p = 0.014 and p = 0.001, respectively). However, no significant differences were found among jaw location, position, implant diameter, or length (p > 0.05).

Conclusions

Robot-assisted dental implant surgery resulted in an accurate placement performance. Further studies are required to clarify long-term effects.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12903-025-06740-6.

Background

Proper implant placement is essential for long-term stability and favorable aesthetic outcomes in implant prosthodontics [1]. Conventional freehand implant surgery relies heavily on the experience of the surgeons, introducing uncertainty into the accuracy of implant placement and leading to serious complications. As a result, computer-assisted implant surgery (CAIS) has been suggested for the visual planning and precise placement of implants in the ideal position [2].

CAIS systems can be divided into static (sCAIS) and dynamic (dCAIS) approaches based on the technology and protocols utilized. Under the sCAIS protocol, a stereolithographic surgical template combined with guided surgical kits is utilized to achieve higher accuracy by physically constraining the drill or implant, whereas dCAIS utilizes optical tracking technology, which enables virtual real-time guidance for the freehand placement of implants [3]. The CAIS technology has demonstrated substantial enhancement in the accuracy of implant placement, yielding remarkable clinical outcomes, particularly in edentulous implant restorations. However, both these approaches have certain drawbacks. Static guides have disadvantages, such as intraoperative guide displacement, non-visibility of the surgical field, and difficulties in water cooling [4, 5]. Furthermore, once the guide template is prepared, modifying the surgical plan becomes impossible during surgery [6]. Additionally, the static guided surgery can become more difficult in the posterior region due to the limited inter-arches space, and although this can be overcome by opening a window on the side wall of the guide ring (“U-shaped guide ring”), the accuracy is reduced [7]. In dCAIS, the three-dimensional position of the drill or implant in the jaw is visible through the screen. The surgeon can make timely adjustments to ensure that implants are placed in the optimal positions. Moreover, planned implant trajectories can be changed intraoperatively to deal with unexpected anatomy [8, 9]. However, it has a steep learning curve, and a long training period is required before it can be applied to patients [10]. Even after comprehensive training, the implantation accuracy relies predominantly on the surgeon’s experience. This is attributed to the inability of dynamic navigation, including surgical templates, to control surgeon’s hand tremors which further impedes improvements in accuracy [11].

Medical robots have been advanced extensively in recent years, improving the accuracy of surgery to the submillimeter level [12, 13], which is attractive for dental implant surgery. Yomi, approved by the FDA in 2017, became the world’s first semiactive robot assistance system to improve the clinical accuracy of dental implant surgery [14]. Zhao introduced the first autonomous dental implant robot by the end of 2017, which is capable of automating the entire process, from implant bed preparation to implant placement, with the surgeon intervening only when necessary [15]. Dental implant robots combine the advantages of the physical constraints of static guide template and real-time feedback of dynamic navigation. More importantly, the entire surgical process can be completed using a mechanical arm, which fundamentally eliminates artificial errors caused by the surgeon’s interventions [16].

The accuracy of robots as novel CAIS methods has been reported in several case reports [11, 17, 18] and in vitro studies [1, 19]; however, reliable accuracy analysis based on a large number of clinical cases is lacking. In addition, the factors that influence accuracy of implant placement remain unknown. In this study, 100 consecutive robot-assisted implant surgeries were analyzed to demonstrate the accuracy performance of the robotic system and identify its influencing factors.

Materials and methods

This study presents a retrospective case series of robot-assisted dental implant surgery for patients with single or multiple missing teeth, which was approved by the ethics committee of Xi’an Jiaotong University Stomatology Hospital, Shaanxi, China (2023-XJKQIEC − 037 − 003), in accordance with the principles outlined in the Declaration of Helsinki of 1975, revised in Fortaleza in 2013. The informed consent was obtained from all patients after a detailed description of the treatment protocol and possible complications prior to the implant surgery. All patients received periodontal treatment to establish optimal oral hygiene one week before operation. This manuscript was prepared conforming to the PROCESS 2020 guidelines for case-series reports.

All surgeries were conducted at the Department of Implant Dentistry at Xi’an Jiaotong University Stomatology Hospital between April 2022 and June 2023. A total of 100 patients underwent implant placement (152 implants) through a flapless procedure by a single surgeon (C.P.Lv) using a commercial image-guided autonomous robotic dental surgery system (Remebot, Beijing Baihui Weikang Technology Co., Ltd.; Beijing, China). As illustrated in Fig. 1b, the Remebot system is primarily composed of surgical navigation software, an optical tracker, and a robotic arm.

Fig. 1.

The treatment protocol, robot system and clinical photos before implant osteotomy. a The treatment protocol of a robot-assisted dental implant surgery. (i) fix the marker in the mouth using acrylic based resin; (ii) CBCT examination for preoperative evaluation; (iii) surgical plan design; (iv) automatic registration and calibration using the optical-system method to achieve higher accuracy and real-time dynamic adjustment; (v) automatic performance of the implant osteotomy and placement under surgeon’s supervision. b The system is consisting of an operating system, a robotic arm, an optical tracker. c-f Representative clinical photos of treatment process i-iv

Participants and protocol

The inclusion criteria were as follows: (1) dentition defect or edentulism; (2) age between 18 and 70 years, regardless of sex; (3) absence of severe bone defects; (4) good oral hygiene and general health; (5) mouth opening of at least 35 mm; and (6) At least two teeth present in the same jaw as the implant site. The exclusion criteria were as follows: (1) pregnancy; (2) uncontrolled systemic diseases (e.g., hypertension, diabetes, bleeding disorders); (3) smoking (≥ 10 cigarettes/day) and alcoholism; (4) contraindication to radiation exposure; (5) allergy to acrylic resin; (6) general contraindications for implant treatment. (7) Insufficient width of keratinized tissue (the width of keratinized tissue around the planed implant needs to be at least 2 mm after flapless procedure).

Based on the manufacturer’s instructions, the treatment protocol involved four steps: (1) first visit, patient screening and general preoperative examination; (2) preoperative preparation, CBCT (Cone Beam Computed Tomography, KaVo Company, Germany) acquisition using a marker for robot identification; (3) intraoperative phase, registration and calibration of the robotic arm (Remebot, China) and patients’ anatomical information, followed by robot-assisted dental implant surgery according to the designed sequence; and (4) postoperative analysis, reacquiring CBCT scan for accuracy assessment (Fig. 1a, c-f).

Preoperative preparation and surgical planning

On the day of surgery, a universal marker was fixed to 2–3 teeth in the residual dentition using a self-cured acrylic resin (ProtempTM, 3 M ESPE, Neuss, Germany. The marker contained an array of three-dimensionally distributed ceramic particles. By recognizing black and white blocks on the marker, the optical tracker (Remebot, China) could ascertain the precise position of each ceramic particle, thereby facilitating the identification of the spatial position of the dentition. Subsequently, the patient underwent CBCT examination using the following machine parameters (tube voltage, 100 kV; tube current and rotation period, 100 mAs; field of view, 16 cm × 10 cm; slice width, 300 μm). Of note, the marker was not touched during the whole process. The acquired CBCT image was then exported in the standard digital imaging and communications (DICOM) format to the Remebot surgical navigation software (Remebot, China). By utilizing three-dimensional reconstruction, artificial intelligence-based structural segmentation, and quantitative analysis provided by the software, the surgeon could observe and analyze the oral structure of the patient. Finally, a virtual implant planning protocol based on the principle of prosthetically driven design was developed, including the implant type and drilling sequence. In this study, Straumann Bone Level Tapered implant (Institut Straumann AG, Switzerland) was used. The diameters and lengths of the implants were selected by the same surgeon based on the patient’s jawbone dimensions.

Intraoperative phase

Following disinfection and sufficient infiltration anesthesia, the surgeon maneuvered the robotic arm near the oral cavity, initiating automatic registration and calibration via the optical-system method. During registration, automatic and manual registration functions were used to identify and extract coordinate information of the ceramic particles in the CBCT image. This information was then matched with the predefined parameters of the ceramic particles on markers to achieve patient registration. Subsequently, the calibration plate was connected to robotic arm in advance and moved approximately 5 cm above the tip of the patient’s nose to register the robotic arm; four positions on the calibration plate were identified and the robotic arm was manual pulled towards the left and right sides of the patient’s mandible to register the last two positions, at this point, all preparations were completed, and the operation was ready to begin. The detailed surgical steps were shown in Fig. 2a, particularly, the surgeon and assistant confirmed the plan for the last time (Fig. 2b), then the drill was connected, and the robotic arm was adjusted to its three-dimensional position and preparation of the implant bed was initiated without incisions (Fig. 2c, d). The surgeon observed the real-time virtual position of the drill or implant in different planes on the screen (Fig. 2e). If necessary, the drilling sequence could be changed intraoperatively. Thereafter, the implant was automatically placed according to preoperative planning (Fig. 2f). When implant placement at one site was complete, the robotic arm moved automatically to the next site according to the plan and the procedure was repeated. Finally, the healing abutment was manually connected to the implant if the insertion torque exceeded 30 Ncm [20, 21]; otherwise, the closure cap was placed, followed by tension-free suturing (local undermining at the implant site might be necessary), and the marker was removed.

Fig. 2.

Detailed surgical steps and postoperative analysis procedures and representative photographs. a Surgical and postoperative analysis protocol. b-f Representative photographs of five surgical steps, including plan confirmation, drill replacement, implant osteotomy, real-time monitoring, and implant placement. g Comparison of implant profiles between the planned (red) and placed (green) implants

Accuracy assessment and analysis

Each patient underwent a postoperative CBCT scan. The preoperative and postoperative CBCT-generated DICOM files were uploaded to the surgery verification software (RemebotDent, China), and the preoperative DICOM file, containing implant planning information, was merged and aligned with the postoperative DICOM file. Subsequently, implant profile extraction was applied to the postoperative images through threshold segmentation and comparison with the implant data. To assess the accuracy of the implant placement, a comparative analysis was conducted using preoperative images. Deviations between the planned and placed implants were automatically calculated (Fig. 2), including coronal, apical, and angular deviations, as previously reported [17, 22](Fig. 3). The distance of deviations was recorded in millimeters, and angle of deviations was measured in degrees.

Fig. 3.

Schematic representation of measurement of deviations between the planned and placed implants

All statistical analyses were performed using SPSS Statistics version 26.0 (SPSS, Armonk, NY, USA). Values are presented as the mean ± standard deviation for normally distributed data or median and interquartile range for continuous variables following non-normal distribution. The Kolmogorov–Smirnov test was applied to test for a normal distribution (α = 0.05). One-way ANOVA (Analysis of Variance) was used for comparison of normally distributed data among multiple groups, and the least significant difference t-test was used for further comparison between two groups. The Kruskal–Wallis test was performed when the distribution was not normal. Multiple hypothesis tests were adjusted using the Bonferroni approach, based on an FWER (Family-wise Error Rate) threshold of 0.05. All tests for significance were two-sided, and significance was defined as values of p < 0.05. All figures were plotted using GraphPad PRISM software (version 9.0; GraphPad Software, Inc., San Diego, US).

Results

Description of the patient population

This retrospective case series evaluated data of 100 patients, including 51 male and 49 female individuals. A total of 152 implants were inserted through flapless surgeries, and only 14 implants underwent submerged healing after suturing due to lack of primary stability (< 30 Ncm). No serious adverse events (e.g., nerve or adjacent tooth injury) were observed. Nonetheless, certain unexpected intraoperative events led to surgical delays, including drill jamming, unforeseen obstructions, and displacement of the optical tracker. After the treatment, some symptoms of postoperative discomfort (slight swelling, loosening of healing abutment, delayed pain of muscle or temporomandibular joint) were also recorded. However, early implant failures (< 1 month) occurred in two cases (two implants) without obvious inflammation. None of the patients underwent guided bone regeneration (GBR) (detailed information was collected in Additional file 1).

Accuracy comparison

As shown in Table 1, the 152 analyzed implants demonstrated coronal deviation of 0.51 ± 0.02 mm, apical deviation of 0.53 ± 0.02 mm, and angular deviation of 1.05 ± 0.05°. The maximum deviations at the cervical, apical, and angular levels were 1.31 mm, 1.33 mm and 3.79°, respectively (Fig. 4). The deviations were further categorized based on the factors influencing the accuracy of implant placement, and the values were statistically compared between the subgroups. Significant differences were observed in angular deviations among different distances from the most-distal natural tooth (first nearest vs. second nearest vs. third nearest, Fig. 5a) and the coronal and apical deviations between among clinical experiences (1st–38th vs. 39th–76th vs. 77th–114th vs. 115th–152nd implant placements, Fig. 5b). No significant differences were observed in other group comparisons (p > 0.05, Fig. 5c-f).

Table 1.

The coronal, apical, and angular deviations following robot-assisted implant placement in the patient case series

All cases		Coronal deviation (mm)					Apical deviation (mm)					Angular deviation (°)
	N	Mean (± SD)	Median	P25 - P75	Min - max	p - Value	Mean (± SD)	Median	P25 - P75	Min - max	p - Value	Mean (± SD)	Median	P25 - P75	Min - max	p - Value
	152	0.51 (± 0.02)	0.49	0.39 to 0.60	0.11 to 1.31		0.53 (± 0.02)	0.52	0.41 to 0.63	0.10 to 1.33		1.05 (± 0.05)	0.91	0.57 to 1.39	0.08 to 3.79
Type of jaw
Maxilla	96	0.50 (± 0.02)	0.49	0.37 to 0.60	0.11 to 0.97	0.924 a	0.53 (± 0.02)	0.52	0.41 to 0.63	0.12 to 0.98	0.933 b	0.99 (± 0.06)	0.90	0.52 to 1.35	0.08 to 2.84	0.317 a
Mandible	56	0.50 (± 0.02)	0.49	0.39 to 0.59	0.12 to 1.31		0.53 (± 0.03)	0.51	0.41 to 0.64	0.10 to 1.33		1.14 (± 0.10)	0.93	0.65 to 1.46	0.08 to 3.79
Region of jaw
Anterior	72	0.53 (± 0.02)	0.51	0.40 to 0.63	0.18 to 0.97	0.078 a	0.55 (± 0.02)	0.53	0.43 to 0.64	0.10 to 0.98	0.067 a	1.01 (± 0.07)	0.94	0.51 to 1.47	0.08 to 2.84	0.614 a
Posterior	80	0.48 (± 0.02)	0.47	0.39 to 0.60	0.11 to 1.31		0.51 (± 0.02)	0.49	0.38 to 0.63	0.12 to 1.33		1.09 (± 0.07)	0.91	0.67 to 1.30	0.19 to 3.79
Implant diameter
< 4.8 mm (3.3 mm, 4.1 mm)	118	0.51 (± 0.02)	0.50	0.40 to 0.60	0.12 to 0.97	0.218 a	0.53 (± 0.02)	0.52	0.42 to 0.63	0.10 to 0.98	0.747 b	1.00 (± 0.05)	0.91	0.54 to 1.39	0.08 to 2.84	0.354 a
= 4.8 mm	34	0.49 (± 0.04)	0.47	0.39 to 0.56	0.11 to 1.31		0.52 (± 0.04)	0.45	0.37 to 0.70	0.12 to 1.33		1.21 (± 0.14)	0.95	0.72 to 1.44	0.33 to 3.79
Implant length
≤ 10 mm (8 mm, 10 mm)	87	0.50 (± 0.02)	0.49	0.39 to 0.59	0.11 to 1.31	0.895 a	0.52 (± 0.02)	0.50	0.40 to 0.66	0.10 to 1.33	0.397 a	1.21 (± 0.07)	0.91	0.65 to 1.34	0.19 to 3.79	0.252 a
> 10 mm (12 mm, 14 mm)	65	0.51 (± 0.02)	0.48	0.38 to 0.60	0.18 to 0.97		0.54 (± 0.02)	50.5	0.42 to 0.63	0.26 to 0.96		0.96 (± 0.07)	0.88	0.47 to 1.44	0.08 to 2.62
Distance from the most-distal natural tooth
1 st nearest	15	0.51 (± 0.07)	0.53	0.42 to 0.59	0.12 to 1.31	0.356 a	0.54 (± 0.07)	0.54	0.45 to 0.60	0.15 to 1.33	0.114 a	0.69 (± 0.09)	0.78	0.29 to 0.88	0.19 to 1.34	0.043 b
2nd nearest	9	0.40 (± 0.05)	0.39	0.32 to 0.50	0.12 to 0.62		0.46 (± 0.06)	0.42	0.41 to 0.60	0.15 to 0.75		1.30 (± 0.32)	0.98	0.68 to 1.67	0.33 to 3.52
3rd nearest	6	0.48 (± 0.02)	0.47	0.45 to 0.52	0.40 to 0.55		0.38 (± 0.04)	0.38	0.29 to 0.47	0.27 to 0.52		1.39 (± 0.32)	1.32	0.72 to 1.91	0.52 to 2.74
Clinical experience
1 st − 38th of the placed implants	38	0.54 (± 0.02)	0.51	0.42 to 0.66	0.19 to 0.87	0.014 a	0.60 (± 0.03)	0.55	0.49 to 0.76	0.25 to 0.98	0.001 a	1.16 (± 0.10)	1.10	0.62 to 1.55	0.19 to 2.84	0.061 a
39th − 76th of the placed implants	38	0.57 (± 0.03)	0.53	0.43 to 0.72	0.12 to 1.31		0.58 (± 0.03)	0.56	0.47 to 0.71	0.20 to 1.33		0.95 (± 0.08)	0.91	0.54 to 1.30	0.08 to 2.44
77th − 114th of the placed implants	38	0.46 (± 0.03)	0.44	0.33 to 0.58	0.18 to 0.96		0.48 (± 0.03)	0.46	0.37 to 0.56	0.10 to 0.98		1.25 (± 0.13)	1.16	0.59 to 1.61	0.12 to 3.79
115th − 152th of the placed implants	38	0.44 (± 0.03)	0.47	0.35 to 0.57	0.11 to 0.79		0.46 (± 0.02)	0.45	0.36 to 0.58	0.12 to 0.74		0.85 (± 0.07)	0.78	0.47 to 1.22	0.18 to 1.88

Bold words represented significant differences (p < 0.05)

^a represented the p value was calculated by Kruskal-Wallis test

^b represented the p value was calculated by one-way ANOVA or T test

Fig. 4.

Accuracy of implant position assessed as coronal deviation in mm, apical deviation in mm, and angular deviation in degrees. The 152 analyzed implants demonstrated coronal deviation of 0.51 ± 0.02 mm, apical deviation of 0.53 ± 0.02 mm, and angular deviation of 1.05 ± 0.05°

Fig. 5.

The 3D deviations of implants placed in different distances from the most-distal natural tooth (a), clinical experience (b), jaws (c), positions (d), implant diameters (e), and lengths (f). *p < 0.05, **p < 0.01, “ns” means no significant difference, multiple hypothesis tests were adjusted via Bonferroni approach based on an FWER threshold of 0.05

For the 30 implants placed in the molar region of distal free-end edentulous spaces, the present study found that as the implant site transitioned from the first nearest location to third nearest location, the angular deviations ranged from 0.69° to 1.39°, showing significant difference (p = 0.043). Multiple comparison results showed that the first nearest site exhibited a smaller angular deviation compared with that of the second and third nearest sites (p = 0.039), whereas no significant difference was observed in the coronal or apical deviations (Fig. 5a).

The coronal deviation of the first 38 placed implants was 0.54 ± 0.02 mm, which increased to 0.57 ± 0.03 mm for the 39th to 76th placed implants, and finally decreased from 0.46 ± 0.03 mm for the 77th to 114th implants to 0.44 ± 0.03 mm for the last 38 placed implants. The corresponding apical deviations were 0.60 ± 0.03 mm, 0.58 ± 0.03 mm, 0.48 ± 0.03 mm, and 0.46 ± 0.02 mm, respectively. Both deviations demonstrated statistically significant differences (p = 0.014 for coronal deviation and p = 0.001 for apical deviation). Additionally, significant improvements in accuracy were achieved after placing half of the 152 implants (Fig. 5b).

Discussion

Ensuring the precise translation of virtual plans to surgical sites is essential for achieving optimal outcomes in dental implant restoration [23]. CAIS technologies have been developed to meet these requirements and achieve prosthetic-driven treatments. Prior understanding of the accuracy and possible influencing factors of each CAIS technique is a prerequisite for selecting an appropriate implant surgery method. However, studies on the accuracy of robot-assisted dental implant surgery are extremely rare and are primarily based on case reports or in vitro model studies. To the best of our knowledge, this is the first clinical case series to analyze the accuracy and identify the influencing factors of a robotic system based on 100 cases. The implant placement accuracy in terms of coronal deviations (0.51 ± 0.02 mm), apical deviations (0.53 ± 0.02 mm), and angular deviations (1.05 ± 0.05°) observed in this study is comparable to that of a previous case series [11]. In that study, 21 patients who underwent robot-assisted surgery for placing 28 implants demonstrated a coronal deviation of 0.53 ± 0.17 mm, an apical deviation of 0.56 ± 0.13 mm, and an angular deviation of 0.79 ± 0.23°. According to 2018 ITI (International Team for Implantology) Consensus Report, the average deviations using surgical template (sCAIS) are 1.2 mm at coronal point, 1.5 mm at apical point, and 3.5° in angulation [24]. Compared with free-hand surgery, which have corresponding deviations of 1.22 ± 0.63 mm, 1.91 ± 1.17 mm, and 7.93 ± 5.56°, respectively [25], the sCAIS technology has improved accuracy and is widely used in clinical settings. However, millimeter-level implantation errors still exist. Navigation-assisted implant surgery is a well-known technique for dCAIS, but the accuracy did not significantly improve in comparison to sCAIS. Previous in vivo studies on navigation systems reported a mean coronal deviations of 0.71–1.36 mm, apical deviations of 0.88–1.83 mm, and angular deviations of 2.26–6.46° [3, 23, 26–31](Fig. 6). These results might be attributed to the intrinsic drawbacks, such as the mechanical errors caused by the cylinder-burr gap [17] or surgeon’s hand tremors [11]. However, with the assistance of a robotic arm, the average distance deviations (coronal and apical aspects) in the present study were successfully decreased to the submillimeter level. This not only ensures the patient safety but also avoids unplanned additional procedures, such as bone grafting or maxillary sinus lift surgery. Interestingly, the angular deviation was reduced to approximately 1°, which may reduce the utilization of angled or customized abutments. However, the maximal deviations observed at the cervical, apical, and angular levels in the present study were 1.31 mm, 1.33 mm, and 3.79°, respectively, suggesting that we should maintain a safe distance from important anatomical structures when planning implant positions, even with systematic training or clinical experience.

Fig. 6.

The accuracy results of this case series compared to previously published studies about navigation-assisted dental implant surgery

All 152 implants were successfully placed without any adverse surgical events, likely due to the reduced physical burden on surgeons. During the surgery, surgeons were mainly responsible for registration, calibration, supervision, drill exchange, implant placement, abutment connection, and wound closure, while the robot performed the rest. This allowed greater attention to the patient’s oral cavity. However, there were some unexpected accidents happened. For example, bone debris left in the implant bed during preparation might have caused drill vibrations. In some cases, the drill could even become stuck, requiring recalibration of the robotic arm, and unnecessarily prolonging the operation time. Robotic surgery was undoubtedly attractive, but unintended blockage or movement of the optical tracker by surgical observers could have led to real-time positioning errors. Moreover, the rCAIS is very difficult in the second molar area due to the limited inter-arch space. This was the reason why only 11 of the involved 152 implants were placed in the second molar area. The maximum deviations were 0.8 mm, 0.78 mm, and 2.74°, respectively. Although clinically acceptable, the procedure was often prolonged due to repeated intraoral verification and frequent reminders for the patient to keep mouth opening. Therefore, we recommend the application of robotic system should be avoided in the second molar area, especially for patients with insufficient mouth opening, sensitive gag reflex and anxiety disorder.

Postoperative complications following robot-assisted surgery are typically mild and manageable, including transient swelling and bleeding. However, different from the results of other studies [11, 17, 32], our study found that two implants occurred early failure without obvious inflammation. Considering both implants underwent non-submerged healing, the reason might be due to the stress on the healing abutment that caused micro-movements and fibrous healing of implants. Since the purpose of this retrospective case series was implantation accuracy, the data within 1 month after surgery were collected retrospectively. Therefore, the analysis of intra/postoperative complications is still needed in the future.

Regarding implantation in free-end arches, as the placement site moved distally, the robotic system demonstrated low levels of coronal (0.40–0.51 mm), apical (0.38–0.54 mm), and angular (0.69–1.39°) deviations. Additionally, no significant differences were observed in the coronal or apical deviation among the different implant sites; however, the angular deviations at the second and third nearest sites were significantly higher than those at the first nearest site. Considering that no relevant study is available for comparison, it can only cautiously speculate that the reason for this might be the limited inter-arch space in the distal region. In contrast, the coronal, apical, and angular deviations of sCAIS in free-end region were reported to be 0.98–1.43 mm, 1.29–1.80 mm, and 3.37–3.88°, respectively [33]. The reason for the poor accuracy may due to the deformation or sinking of the guide template during drilling [34, 35]. The performance of dynamic navigation in free-end implantations has rarely been investigated. Schnutenhaus et al. [36] placed 120 implants in regions 45 and 47 (FDI dental numbering system) using a dynamic navigation system. They reported mean coronal, apical, and angular deviations were 1.52 mm, 1.81 mm, and 2.87° in region 45, and 1.54 mm, 1.77 mm, and 2.89° in region 47. Although it has demonstrated improved accuracy than sCAIS, it still appears inferior to robotic systems. Considering that they possess similar optical tracking technology, the improved accuracy may be attributed to the use of a robotic arm [11]. The behavior in free-end implantation is particularly attractive to patients with edentulism [37].

In this study, the surgeon’s clinical experience with robot-assisted dental implant surgery significantly influenced the accuracy of the implant placements at the coronal and apical levels, even though the surgeon had undergone long-term training before its application on patients. This result is consistent with that of the EAO (European Association for Osseointegration) Consensus Conference (2012) warning against compensating for the lack of training with CAIS technology [38]. Similar findings were observed in subsequent studies, in which novice surgeons without much experience in implant surgeries exhibited significantly poorer accuracy than that of experienced surgeons, regardless of whether a guide template [39, 40] or dynamic navigation [28] was used. In present study, the robot system was doing the same work theoretically, there was no machine learning used for the surgical execution. The improvement in accuracy may be attributed to the team’s clinical experience (more precisely), including not only the surgeon but also nurses, engineers, and radiologists. For example, with the accumulation of experience, (1) surgeons will know the importance to timely remove bone scraps to avoid drill vibration. Moreover, the selection of long and short Straumann drills in the preoperative planning stage can be more appropriate; (2) Nurses are trained to control the number and movement of people in operation room to prevent disruption of the optical tracking system; (3) Engineers, as non-medical workers, will better understand the surgical process and promptly adjust the position of robotic system for facilitating surgical field observation; (4) Radiologists will try to prepare CBCT in advance to reduce patient examination time and set appropriate parameters to reduce radiation dose. However, it remains unclear which specific clinical modification that led to improved accuracy at later stages. Moreover, it is worth noting that the observed trend in accuracy improvement did not necessarily begin with the placement of the 76th implant, which was due to the quartile-based categorization.

Other possible influencing factors, including the jaw (maxilla vs. mandible), as well as position (anterior vs. posterior), implant diameter (< 4.8 mm vs. 4.8 mm), and implant length (≤ 10 mm vs. >10 mm) of the implants had not observed any significant impact on the final 3D deviations. These results are consistent with those of a previous case series [11], in which the robot system demonstrated a stable and accurate placement performance across a reasonable range of implant sites (jaws and positions) and implant sizes (diameter and length), except for apical deviations in different jaws (p = 0.0387). The robust outcome of the robotic implantation accuracy in this study was beyond our expectations. Different jaws and positions have been reported to have a non-negligible impact on other CAIS technologies. For instance, the accuracy of CAIS technology in the mandible or posterior region is worse than that in the maxilla or anterior region due to the intraoperative movement of the mandible [11] or the obstruction of the antagonist tooth in the posterior region [41]. This is consistent with the findings of Liu et al. [42], although the underlying reason remains unclear. We cautiously propose two possible explanations: (1) the real-time and rapid tracking and adaptation of the mandibular position by the optical tracker and robotic arm [43]; and (2) the elimination of finger rests on the dentition by the surgeon, potentially enhancing mandibular stability.

Whether the implant diameter and length affect the placement accuracy in the sCAIS method remains controversial. Kim et al. [44] recently reported a significant difference in the coronal and apical deviations during implant placement based on implant diameter and length. Similarly, a template-guided implant placement study also revealed that implant insertion with a length of 8–9 mm resulted in significantly higher precision than that of insertion with lengths of 10–11 and 12–13 mm [45]. However, other studies have suggested that the diameter and length have no significant impact on the placement accuracy [46, 47]. These discrepancies could be related to various factors, such as the implant system, template design software, in vivo or in vitro studies, and CBCT quality. In this study, the implant diameter or length did not affect the accuracy of robotic implantation, which is consistent with the results of a previous study [11]. This may be due to the control of the drilling direction by the robot system. However, further clinical studies are needed to confirm these findings. The observed high level of accuracy may be attributed to three technical aspects: (i) special robotic arm enabling automatic correction of drill post and reliable maintenance of drill direction compared to manual arm [48]; (ii) the visual servo-loop of the robot to compensate for small patient movements [49]; and (iii) minimal registration error (< 0.1 mm) after optical tracker-to-robot registration [13]. In addition to these technical factors, effective cooperation between well-trained surgeons, professional manufacturer engineers, and robot systems, as well as a timely postoperative review, may be significant factors.

This study also has several limitations. Firstly, no cases involved GBR or maxillary sinus lifting, and only one implant and robotic system were used. Secondly, the classification standards of long/short [50, 51]and wide/narrow implants [52, 53] were based on previous studies. Finally, the multicenter clinical studies are still needed in the future.

Conclusions

Within the limitations of this case series, robot-assisted dental implant placement demonstrated excellent accuracy without the need for additional procedures, such as guided bone regeneration. The data appear promising when compared to both static and dynamic navigation; however, further studies are necessary to ensure successful osseointegration after robot-assisted dental implant surgery in the future.

Abbreviations

CAIS

Computer-assisted implant surgery

DICOM

Digital imaging and communications in medicine

FWER

Family-wise Error Rate

ANOVA

Analysis of variance

CBCT

Cone beam computed tomography

EAO

European association for osseointegration

ITI

International team for implantology

3D

Three-dimensional

Author’s contributions

Chengpeng Lv# and Bowen Qin# should be considered joint first author.Chengpeng Lv: Conceptualization; Writing - original draft; Methodology; Data curation; Investigation. Bowen Qin: Methodology; Data curation; Investigation; Formal analysis; Writing - review & editing. Longlong He: Methodology; Data curation; Investigation. Boya Xu: Methodology; Supervision; Data curation; Investigation. Yiqun Wu: Writing - review & editing; Project administration; Methodology; Supervision. Zhe Li: Conceptualization; Writing - original draft; Methodology; Supervision; Writing - review & editing; Project administration; Formal analysis. Liangzhi Du: Conceptualization; Writing - review & editing; Methodology; Supervision; Project administration; Resources.

Funding

This study was supported by National Key Research and Development Program of China (2019YFB1302204); National Natural Science Foundation of China (82301155); Key R&D Program of Shaanxi Province (2023- YBSF-220; 2023-YBSF-193).

Data availability

Available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the ethics committee of the Xi’an Jiaotong University Stomatology Hospital, Shaanxi, China (2023-XJKQIEC − 037 − 003). ‘Informed consent’ was obtained from patients and gave us the right to utilize their data. Clinical trial number: not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.