Feasibility of Mandibular Distraction Osteogenesis Using Mixed Reality-based Dynamic Navigation: A Preclinical Study

Abstract

Background

Mandibular distraction osteogenesis (MDO) is critical for correcting mandibular hypoplasia. Mixed reality (MR) is an emerging surgical-navigation technology that can overcome the limitations of traditional methods. However, evidence regarding the benefits of MR navigation in MDO is limited. Herein, we developed a dynamic MR-based navigation system and compared its performance to techniques using surgical guides (SGs) as well as freehand (FH) and traditional navigation (TN) approaches.

Methods

An adjacent-display MR navigation system integrating optical tracking and a patient-specific registration device was developed for real-time dynamic guidance. Thirty mandibular three-dimensional (3D)-printed models were allocated to three groups (n = 10 each): MR, SG, and FH. Thirty-two Beagle dogs were randomised into four groups (n = 8 each): MR, SG, FH, and TN. The feasibility of the MR system was evaluated in 10 human cadaver heads. All procedures were performed by the same surgical team. The outcomes included osteotomy accuracy, distraction-vector deviation, distractor placement precision, intraoperative corrections, and operative time.

Results

In the model experiments, the MR group showed significantly higher distractor and osteotomy accuracy than the FH group and similar precision as the SG group. In animal studies, the MR approach reduced errors in distractor placement, distraction vector, and osteotomy in comparison with the FH approach, and outperformed the SG-based approach in terms of distractor positioning and osteotomy accuracy. In comparison with TN, the MR approach excelled in distractor positioning, osteotomy accuracy, and reduced operative time. Cadaver experiments revealed sagittal and transverse distractor errors of 1.99° ± 0.67° and 1.63° ± 0.88°, respectively, distraction-vector deviation of 2.62° ± 0.38°, and osteotomy angle deviation of 1.74° ± 0.50°. The average operative time was 85.23 ± 13.04 min; distractor positioning required 54.08 ± 15.71 min and the number of intraoperative corrections was 6.80 ± 1.87.

Conclusions

The dynamic MR navigation system achieved better accuracy than the FH approach and showed specific advantages over the TN and SG-based approaches, indicating its potential clinical applicability for precise MDO.

Trial Registration

This study was registered with the Chinese Clinical Trial Registry (ChiCTR2500098863).

Introduction

Distraction osteogenesis (DO) is a well-established surgical technique wherein gradual mechanical traction after osteotomy stimulates new bone formation, remodelling, and soft tissue adaptation. DO has become the preferred treatment for severe mandibular hypoplasia, particularly in reconstructive procedures, because it allows simultaneous correction of bone deficiency and soft tissue adaptation.¹ Despite these advantages, mandibular distraction osteogenesis (MDO) is technically demanding and associated with complications such as injury to dental germs, damage to the inferior alveolar nerve, and anterior open bite.² These problems are often related to inaccuracies in distractor placement, osteotomy design, or pin configuration.³ Since the distraction vector is determined directly by distractor positioning, even small errors may result in mandibular asymmetry or malocclusion.⁴ Therefore, accurate preoperative planning and precise intraoperative execution are essential for optimising outcomes and reducing complications.

Various approaches have been developed to improve accuracy in MDO, including the use of pre-bent distractors, computer-assisted surgical navigation (CASN), and patient-specific surgical guides (SGs).^5,6 Virtual surgical planning (VSP) and computer-aided design (CAD) can facilitate accurate preoperative design of osteotomy lines and distractor positioning. However, VSP alone cannot guide intraoperative procedures, and the transfer of pre-bent distractors from physical models to the operating field is heavily dependent on the surgeon’s experience level. Consequently, discrepancies in this transfer process can introduce cumulative angular errors and cause deviations in distractor and osteotomy placement, thereby affecting surgical precision.⁷ SGs can provide anatomical references, but their accuracy depends on perfect bone-surface adaptation, and they have shown limited utility for evaluating proximity to critical anatomical structures.⁸ In contrast, conventional CASN systems allow real-time three-dimensional (3D) guidance, but they require the surgeon to switch visual attention between the operative field and an external screen, potentially increasing cognitive workload and decreasing surgical efficiency.⁹

Mixed reality (MR) has been recently introduced as an alternative approach to oral surgical navigation.^10,11 By integrating virtual information into the operative environment, MR can provide more intuitive intraoperative guidance.^12,13 The use of optical see-through head-mounted displays (OST-HMDs), such as the Microsoft HoloLens, has been explored in oral and maxillofacial surgery.14, 15, 16 Nevertheless, overlay-based MR systems show some limitations. Persistent holographic overlays may obscure the operative field, disrupt the surgical workflow, and pose barriers to routine clinical adoption in complex procedures.¹⁷ Moreover, the spatial disorientation and misalignment caused by the conversion of 3D planning data into two-dimensional (2D) screen views may interfere with the surgeon’s natural perspective, reducing intuitive control and accuracy.¹⁸

The existing studies on the use of MR in MDO are limited and have focused mainly on the feasibility of MR-assisted osteotomy without attempting comparative evaluations with other surgical techniques or detailed assessments of distraction-vector accuracy, which are critical for ensuring postoperative symmetry.^19,20 Although our previous studies^7,21 with HoloLens 2 demonstrated the feasibility of this approach in model-based distractor placement, the findings showed that several technical challenges were unresolved, including persistent visual obstruction, limited visualisation of adjacent anatomical structures, and instrument tracking. Moreover, most of the existing MR systems used in MDO function as static display tools, and their inability to dynamically track surgical instruments restricts their intraoperative applicability. Further studies are needed to evaluate the advantages of MR in this domain.

Given these limitations, the present study attempted to develop a dynamic MR navigation system designed for MDO. By adopting an adjacent-display approach, the system separated virtual information from the surgical field, thereby avoiding the visual obstruction caused by traditional overlays. Importantly, the system allowed real-time dynamic tracking of surgical instruments, providing continuous intraoperative feedback beyond static anatomical visualisation. This design allowed surgeons to visualise distractor positioning, osteotomy lines, and adjacent anatomical structures while monitoring their operative tools. The accuracy and feasibility of this system were comprehensively evaluated through multilevel preclinical experiments with patient-specific mandibular models, animal subjects, and human cadaveric specimens, and the findings were directly compared to the corresponding findings for other surgical techniques. The ultimate goal of this research was to advance surgical navigation in MDO by introducing a more accurate, reliable, and adaptable system, which could potentially support safer procedures and yield improved outcomes in maxillofacial practice.

Materials and techniques

Study subjects

The cohort of 30 anatomically accurate, fully dentate, 1:1 scale mandibular representation models was meticulously reconstructed from the computed tomography (CT) scans of 10 individuals diagnosed as showing mandibular hypoplasia, with three models for each patient. The inclusion criteria were an intact mandibular anatomy and age ≥18 years. The animal group included 32 adult healthy Beagle dogs (age, 1-2 years; weight, 10-15 kg) obtained from the Laboratory Animal Center of Guangxi Medical University, which has been accredited for laboratory animal research and care. The cadaveric group consisted of 10 human cadaver heads (preserved in 10% formalin) obtained from the Department of Anatomy, Guangxi Medical University. All cadaveric specimens had intact facial soft tissues, mandibular structures, and dentition.

MR system for MDO

The MR system designed for multi-domain operations consisted of a near-infrared optical-tracking device (OP-M620; Aimooe), Microsoft HoloLens 2 (Microsoft), calibration markers, and a computational workstation (Figure 1A). A patient-specific automated reference device, which was based on a custom occlusal splint, linked the mandible to the tracking system. The intraoral component was securely attached to the dentition, whereas the extraoral section consisted of four infrared reflective markers for precise optical tracking. A computer numerical control (CNC)-fabricated rigid frame with six markers allowed accurate localisation of surgical instruments, with pivot calibration ensuring submillimetre precision (Figure 1B). The optical tracker transmitted real-time positional data to the computer for processing, whereas HoloLens 2 rendered a 3D holographic mandible for intraoperative navigation (Figure 2). The software platform, which was developed with Unity 3D and Visual Studio, provided an interactive visualisation interface.

Fig. 1.

Main hardware devices used. A, Optical locator, B, HoloLens 2 and surgical instruments.

Fig. 2.

Principle of the MR-guided navigation system for MDO.

The MR-guided MDO workflow consisted of four steps: (1) hologram adjustment, (2) navigated drilling, (3) osteotomy, and (4) distractor fixation. Surgeons used the HoloLens 2 to visualise anatomical landmarks (Figure 3A), including the mandible, inferior alveolar nerve, and dental roots, along with osteotomy lines and fixation points. Hand gestures allowed dynamic adjustment of hologram position, scale, and rotation (Figure 3B). The system offered two display modes: a drilling mode showing key targets (Figure 3C) and an osteotomy mode with a translucent mandibular view for real-time tracking of adjacent structures (Figure 3D; see Supplementary Video 1).

Fig. 3.

MR-guided visualization and interaction during MDO. A, Surgeon using MR navigation. B, Navigation image adjustment. C, MR-guided drilling. D, MR-assisted osteotomy (Images B–D captured via HoloLens 2 camera.)

Experimental design

In this study, we aimed to validate the accuracy and intraoperative feasibility of the MR-based digital workflow for MDO (Figure 4). To this end, we employed a three-stage preclinical validation framework consisting of a phantom model, live animals, and cadaveric specimens to assess the MR navigation system under progressively increasing levels of anatomical and procedural complexity (Figure 5). The model experiment allowed controlled, repeatable testing of spatial accuracy under simplified anatomical conditions. Subsequently, the animal experiments introduced physiological variables such as soft tissue interference, bleeding, and respiratory motion, thereby allowing evaluation of intraoperative robustness. Both stages were independently designed and analysed without cross-platform comparisons to avoid confounding. Finally, the cadaveric phase, although limited by specimen availability, simulated near-clinical conditions and served as a translational feasibility assessment that supplemented the findings from the earlier stages.

Fig. 4.

Workflow roadmap of MR-assisted navigation for MDO.

Fig. 5.

Flowchart of the experimental design.

Surgical planning and procedure

Preoperative mandibular 3D models were integrated into Proplan CMF 3.0.1 software (Materialise) for VSP. The osteotomy lines were meticulously delineated, with body osteotomies generally situated distal to the last molar and ramus osteotomies strategically positioned superior to the mandibular lingula to prevent potential injury to the inferior alveolar nerve. Customised surgical templates were subsequently designed and fabricated using a 3D printer (Supplementary Figure 1). Optical navigation information was specifically designed to guide screw positioning and determination of the osteotomy site (Supplementary Figure 2). The MR-based surgical-navigation information was designed to include the osteotomy line and screw hole positions, along with visualisation of critical anatomical structures such as the inferior alveolar nerve, vascular bundle, and dental roots (Figure 6).

Fig. 6.

Design of the patient-specific automatic registration device and MR-based surgical navigation information. A and B, Design of the customised automatic registration device. C, Preoperative planning of osteotomy line and screw hole positions. D, Visualisation of anatomical structures (inferior alveolar nerve, vascular bundle, and dental roots) in transparent MR view.

Four surgical methods were used (Figure 7): the freehand (FH) approach relied on preoperative planning and experience; traditional navigation (TN) employed optical navigation through anatomical landmarks; the surgical guide (SG) approach employed specific templates; and the MR approach employed HoloLens 2 for navigation. All MDO procedures employed bilateral distractor placement.

Fig. 7.

Surgical group assignments. A, FH Group, B, TN Group, C, SG Group, and D, MR Group.

In the mandibular model experiment, 30 mandibular models were randomly allocated into MR (Figure 8A and B) (Supplementary Video 2), SG, and FH groups (n = 10 each). Each patient’s individual surgical plan (10 different plans in total) was consistently replicated across these groups to ensure comparability. Postoperative data were acquired by cone-beam computed tomography (CBCT).

Fig. 8.

MR-guided visualisation during MDO in both model and animal experiments. A and B, MR visualisation of screw hole and osteotomy point positions. C and D, Switchable MR display modes illustrating screw hole positioning and osteotomy navigation during surgery (Images captured via HoloLens 2 integrated camera).

In the animal experiment, 32 adult Beagle dogs were randomised into MR (Figure 8C and D) (Supplementary Video 3), SG, TN, and FH groups (n = 8 per group). All surgeries were conducted using an extraoral approach. Postoperative evaluations were performed by CBCT imaging. After completion of the surgery, euthanasia was humanely administered by an intravenous pentobarbital sodium overdose (150 mg/kg).

The human cadaveric study involved 10 cadaveric head specimens. MDO procedures were conducted using intraoral approaches guided by the MR navigation system (see Supplementary Video 1). Postoperative imaging data were collected by CT. Upon completion of the study, the specimens were cleaned in accordance with ethical standards and returned to the anatomy department for proper disposal.

All procedures were performed by the same team with extensive experience in MDO; the team consisted of one senior surgeon with over 15 years of experience, two attending surgeons with more than 5 years of experience each, and two residents. To minimise potential recall bias and ensure procedural consistency, the surgical procedures in each group were conducted at 15-day intervals. All surgeons received standardised training on HoloLens 2 and surgical instruments and underwent individual eye-tracking calibration before surgery.

Quantitative parameters

Postoperative data in the Digital Imaging and Communications in Medicine (DICOM) format were imported into 3-Matic 13.0 and aligned with preoperative models using uniform thresholding, point-to-point, and global registration. For spatial consistency,²² all mandibular models were aligned to a Cartesian coordinate system, with the XY plane representing the horizontal orientation, the YZ plane representing the mid-sagittal orientation, and the XZ plane representing the coronal orientation (Figure 9A).

Fig. 9.

Schematic illustration of measurement parameters. A, Spatial orientation, B, sagittal-plane angular, C, distractor positioning, and D, osteotomy location.

The distraction-vector angle error was defined as the sagittal-plane difference between the planned and actual vectors (Figure 9B). In mandibular model experiments, because of the near-perpendicular distractor orientation, only sagittal errors were analysed. The positioning error was calculated as the centroid deviation of the four distractor pinholes (Figure 9C). Osteotomy points A and B were defined by intersections with the superior and inferior mandibular borders in the sagittal plane, with the angle error measured between the planned and actual vectors (Figure 9D). Intraoperative challenges, such as bleeding, repositioning, or drilling deviations, were recorded. Operative time included the duration of the procedure and the time required for distractor placement.

Statistical analysis

Statistical analyses were performed using SPSS version 26.0 (IBM Corp). Data normality was assessed with the Shapiro–Wilk test. For normally distributed variables, one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test was performed. For variables that did not show a normal distribution, the Kruskal–Wallis test with Dunn’s post-hoc test was used. Descriptive statistics were calculated for all study variables. Depending on the distribution, the results were presented as mean ± standard deviation (SD) or as median with interquartile range (P25, P75). The analysed variables included osteotomy accuracy, distraction-vector deviation, distractor placement precision, intraoperative corrections, and operative time. A two-sided P value < .05 was considered statistically significant.

Results

Mandibular model experiments

Table 1 summarises the accuracy outcomes of the three groups in the model experiments. Overall, the MR group consistently demonstrated the lowest errors for all parameters, and it showed significantly lower distractor angular deviations, sagittal-plane discrepancies, centroid deviations, and osteotomy positioning and angular errors than the FH group (all P < .001). In contrast, the MR and SG groups showed no significant differences across the evaluated indices. These findings indicated that MR navigation achieved comparable accuracy to the SG-based approach and substantially outperformed FH procedures.

Table 1.

Surgical results of the model experiment.

	MR	SG	FH	P value	Multiple comparisons adjusted P value
	N = 20	N = 20	N = 20		MR vs SG	MR vs FH	SG vs FH
Distractor angularK (°)	2.24 ± 0.76	2.58 ± 1.54	5.35	<.0001	>.9999	.0001	<.0001
			(3.76, 8.40)	⁎⁎⁎	Ns	⁎⁎⁎	⁎⁎⁎
Sagittal planeK (°)	2.14 ± 0.62	1.32	4.90	<.0001	.3364	.0006	<.0001
		(0.75, 2.37)	(3.12, 9.32)	⁎⁎⁎	Ns	⁎⁎	⁎⁎⁎
Distractor centroidK(mm)	0.68	0.61	2.54 ± 1.59	<.0001	>.9999	.0001	<.0001
	(0.46, 0.98)	(0.30, 1.22)		⁎⁎⁎	Ns	⁎⁎	⁎⁎⁎
Osteotomy pointAK(mm)	0.87 ± 0.39	1.12	3.51 ± 2.24	<.0001	.4348	<.0001	.0035
		(0.83, 1.40)		⁎⁎⁎	Ns	⁎⁎⁎	⁎
Osteotomy pointBA(mm)	0.85 ± 0.33	1.32 ± 0.83	4.70 ± 2.29	<.0001	.5521	<.0001	<.0001
				⁎⁎⁎	Ns	⁎⁎⁎	⁎⁎⁎
Osteotomy angularA (°)	1.17 ± 0.40	1.92 ± 0.99	5.26 ± 2.50	<.0001	.2941	<.0001	<.0001
				⁎⁎⁎	Ns	⁎⁎⁎	⁎⁎⁎

FH, freehand; MR, mixed reality; ns, not significant; SG, surgical guides.

N = 20 (N represents the total number of measurements: 10 models × 2 measurements per model, one for each side).

Data are shown as mean ± standard deviation, while data not following a normal distribution are expressed as median (P25, P75).

A represents the one-way ANOVA for comparing the three groups, followed by Tukey's multiple comparisons test for pairwise comparisons.

K represents the Kruskal-Wallis test for comparing the three groups, followed by Dunn's multiple comparisons test for pairwise comparisons.

^⁎P < .01.

^⁎⁎P < .001.

^⁎⁎⁎P < .0001.

Animal experiments

Table 2 summarises the measurement results of surgical errors in the canine MDO experiments. The MR group generally exhibited the highest accuracy across most parameters, showing significantly lower distractor angular errors, sagittal-plane deviations, centroid deviations, and osteotomy positioning and angular errors than the FH group (all P < .001). MR navigation also outperformed the SG-based and TN approaches in certain indices, such as distractor centroid deviation and osteotomy point localisation (Figure 10).

Table 2.

Measurement results of surgical errors in canine MDO.

	MR	SG	TN	FH	P value	Multiple comparisons adjusted P value
	N = 16	N = 16	N = 16	N = 16		MR vs SG	MR vs FH	MR vs TN
Distractor angularA (°)	2.24 ± 1.33	3.58 ± 1.73	3.24 ± 1.74	5.57 ± 3.15	<.0001	.2856	.0002	.5359
					⁎⁎⁎⁎	Ns	⁎⁎⁎	Ns
Sagittal planeA (°)	1.22 ± 0.99	2.46 ± 1.45	2.77 ± 1.58	4.94 ± 2.69	.0001	.2160	<.0001	.0807
					⁎⁎⁎⁎	ns	⁎⁎⁎⁎	ns
Transverse planeK (°)	1.54 ± 1.46	2.24 ± 1.47	2.66 ± 1.62	1.96 (1.23, 5.45)	.2109	>.9999	.6766	.2646
					Ns	Ns	ns	ns
Distractor centroidK(mm)	0.63 (0.51, 1.33)	1.88 (1.02, 3.60)	1.32 ± 0.62	2.03 (1.51, 2.74)	.0003	.0024	.0007	.0007
					⁎⁎⁎	⁎⁎	⁎⁎⁎	⁎⁎⁎
Osteotomy pointAA(mm)	1.05 ± 0.54	2.34 ± 0.88	1.92 ± 0.95	2.51 ± 1.22	.0001	.0014	<.0001	.0530
					⁎⁎⁎⁎	⁎⁎	⁎⁎⁎⁎	Ns
Osteotomy pointBA(mm)	1.04 ± 0.45	2.41 ± 0.95	1.84 ± 0.83	2.65 ± 1.23	.0001	.0007	<.0001	.0425
					⁎⁎⁎⁎	⁎⁎⁎	⁎⁎⁎⁎	*
Osteotomy angularA (°)	1.22 ± 0.49	1.96 ± 0.94	2.13 ± 0.88	4.47 ± 1.24	.0001	.1240	<.0001	.0394
					⁎⁎⁎⁎	Ns	⁎⁎⁎⁎	*

FH, freehand; MR, mixed reality; ns, not significant; SG, surgical guides; TN, traditional navigation.

Data are shown as mean ± standard deviation, while data not following a normal distribution are expressed as median (P25, P75).

N = 16 (N represents the total number of measurements: 8 dogs × 2 measurements per dog, one for each side).

A represents the one-way ANOVA for comparing the four groups, followed by Tukey's multiple comparisons test for pairwise comparisons.

K represents the Kruskal-Wallis test for comparing the four groups, followed by Dunn's multiple comparisons test for pairwise comparisons.

^⁎P < .05.

^⁎⁎P < .01.

^⁎⁎⁎P < .001.

^⁎⁎⁎⁎P < .0001.

Fig. 10.

Comparison of surgical accuracy and efficiency among four experimental groups in the canine model. A, Centroid deviation of distractor placement. B, Angular deviation of the distraction vector. C, Sagittal plane deviation of the distraction vector. D, Comparison of total operative time. *P < .05; **P < .01; ***P < .001; ****P < .0001. FH, freehand; MR, mixed reality; SG, surgical guides; TN, traditional navigation; ns, not significant.

Table 3 summarizes the intraoperative performance indicators in the canine MDO experiments. The MR group exhibited significantly shorter total operative time and distractor placement time compared to the TN group. However, no significant differences were observed between the MR group and the SG group or FH group. In terms of intraoperative setbacks, no statistically significant differences were found between the MR group and the other groups (SG, FH, TN).

Table 3.

Intraoperative performance indicators measurement results.

	MR	SG	FH	TN	P value	Multiple comparisons adjusted P value
	N = 8	N = 8	N = 8	N = 8		MR vs SG	MR vs FH	MR vs TN
Total operative timeA	108.90 ± 15.89	96.13 ± 12.72	109.90 ± 12.63	136.90 ± 15.40	<.0001	.2988	.9990	.0027
					⁎⁎⁎	ns	ns	⁎⁎
Distractor placement timeA(min)	71.00 ± 16.51	54.63 ± 12.55	75.38 ± 19.39	100.60 ± 19.47	<.0001	.2137	.9502	.0061
					⁎⁎⁎	ns	ns	⁎⁎
Intraoperative setbacksA(n)	5.63 ± 3.16	11.63 ± 4.27	6.50 ± 4.60	11.50 ± 5.58	.0163	.0565	.9795	.0637
					*	ns	ns	ns

FH, freehand; MR, mixed reality; ns, not significant; SG, surgical guides; TN, traditional navigation.

N = 8 (N represents the number of dogs).

Data are shown as mean ± SD.

A represents the one-way ANOVA for comparing the four groups, followed by Tukey's multiple comparisons test for pairwise comparisons.

^⁎P < .05.

^⁎⁎P < .01.

^⁎⁎⁎P < .0001.

Cadaveric experiments

In the cadaveric specimens, the distractor angle error was 2.62° ± 0.38°; sagittal-plane error was 1.99° ± 0.67°; transverse plane error was 1.63° ± 0.88°; and centroid deviation was 0.87 mm (0.61, 1.08 mm). The osteotomy positioning error was 0.82 ± 0.35 mm for point A and 0.85 mm (0.55, 1.22 mm) for point B, whereas the osteotomy angle error was 1.74° ± 0.50°. The total operative time was 85.23 ± 13.04 min; distractor placement time was 54.08 ± 15.71 minutes; and the number of intraoperative corrective manoeuvres was 6.80 ± 1.87 (Table 4).

Table 4.

Measurement results of MR-guided MDO in cadaveric specimens.

Parameters	Distractor angular (°)	Sagittal plane (°)	Transverse plane (°)	Distractor centroid(mm)	Osteotomy point A(mm)	Osteotomy point B(mm)	Osteotomy angular (°)	Total operative time (min)	Distractor placement time (min)	Intraoperative setbacks (n)
Value	2.62 ± 0.38	1.99 ± 0.67	1.63 ± 0.88	0.87 (0.61, 1.08)	0.82 ± 0.35	0.85 (0.55, 1.22)	1.74 ± 0.50	85.23 ± 13.04	54.08 ± 15.71	6.80 ± 1.87

(N = 10, N represents the number of cadaveric specimens).

Data are shown as mean ± SD or median (P25, P75) as appropriate.

Discussion

DO is the principal therapeutic approach for severe mandibular hypoplasia.²³ However, traditional DO shows lower accuracy than orthognathic surgery, frequently necessitating additional corrective interventions and thereby limiting the broader clinical adoption of this technique.^24,25 To address these limitations, we developed and evaluated a novel surgical-navigation system that integrated automatic registration with patient-specific occlusal splints, MR, and optical-tracking technologies. Using an adjacent display for holographic visualisation, the system provided real-time, independent visualisation to enhance depth perception and potentially reduce intraoperative complications, shorten the operative time, and yield improvements in surgical outcomes.

MR primarily enhances surgical visualisation by overlaying virtual images onto the operative field.²⁶ Although multiple studies on MR have reported favourable outcomes, clinical translation of this technology remains limited, particularly due to inconsistencies in depth perception in confined surgical environments.²⁷ In craniofacial procedures, restricted access, dynamic soft tissues, and visual obstructions from instruments often lead to holographic oversaturation. This form of oversaturation causes visual-convergence mismatches, cognitive overload, and poor localisation of instruments near critical neurovascular structures, thereby increasing operator fatigue and procedural error rates. Kihara²⁸ showed that in comparison with traditional overlays, cross-sectional 2D navigation using MR-HMDs reduces the risk of over-preparation in dental procedures. However, 2D displays show limitations in complex procedures such as osteotomies and distractor placement, where the lack of spatial depth compromises accuracy and workflow efficiency.

In this study, we introduced a holographic navigation system that combines high-fidelity mapping with an adjacent-display scheme, spatially separating the intraoperative guidance information from the surgical field. Unlike commonly used overlay-based MR methods, which project navigation data directly onto the operative area, this interface independently rendered real-time surgical instrument positions and anatomical details, thereby reducing visual interference from hologram–tissue overlap. In the animal experiments, the system yielded shorter operative times and necessitated fewer intraoperative adjustments, advantages that can be attributed to its decoupled display and intuitive human–machine interaction design. Using this system, surgeons could dynamically adjust holographic elements such as position, scale, rotation, and visibility in real time, which minimised visual obstruction and reduced interference with surgical targets. Moreover, the real-time instrument tracking afforded by the system allowed continuous intraoperative monitoring of critical anatomical structures such as neurovascular bundles and dental roots, potentially reducing the risk of iatrogenic injury during osteotomy and distractor placement. To further optimise the osteotomy procedures, the system included a holographic transparency mode, which enabled surgeons to reference preoperative osteotomy planes while simultaneously tracking virtual saw trajectories in real time, thereby avoiding the gaze diversion commonly required in conventional navigation systems and overlay-based MR displays. By overcoming these limitations, the proposed system represents an innovative approach that may enhance both surgical efficiency and safety in MDO.

Furthermore, the automatic registration device for spatial positioning proposed in this study served as a practical method for integrating adjacent-display MR visualisation and optical-tracking technology. By establishing the mandible–device spatial relationship preoperatively, this device allowed direct intraoperative registration without additional CT-based reference tools or anatomical landmarks. The device was rapidly and non-invasively fixed during surgery, and it employed a simple design requiring only minimal customisation. In the animal experiment, the MR group showed a shorter operative time and required fewer intraoperative corrections than the TN group, supporting the device’s ability to streamline surgical workflow and enhance intraoperative efficiency.

To the best of our knowledge, this study is the first to compare MDO performed using an MR-based navigation system with that performed using SG-based, TN, and FH approaches. The analysis included a multilevel comparison encompassing mandibular models, animal experiments, and cadaveric specimens. The key parameters evaluated in the study included osteotomy angle deviation, distraction-vector error, distractor positioning accuracy, and operative efficiency. Precise sagittal control of the distraction vectors in bilateral mandibular distraction is essential, and it directly influences the direction of mandibular advancement.²⁹ In this regard, improper control, eg, counterclockwise rotation, may cause posterior open bite, whereas clockwise rotation may cause anterior open bite. Nonparallel bilateral distraction vectors result in coronal rotation and occlusal plane changes, whereas horizontal plane inaccuracies cause mandibular asymmetry. In experiments with the mandibular model, the MR group showed a significantly lower distraction-vector angle error (2.24 ± 0.76), sagittal-plane error (2.14 ± 0.62), distractor positional error (0.68 [0.46, 0.98] mm), and osteotomy angle deviation in comparison with the FH controls, without showing significant differences in comparison with the SG group. Similarly, in animal experiments, the MR group showed superior accuracy and surgical efficiency in comparison with the FH group, with the distractor centroid positioning error (0.63 [0.51, 1.33] mm) significantly outperforming all other groups. These results reflect the robustness of the MR navigation system, showing consistent accuracy across model and animal tests without causing a notable increase in errors.

Considering the clinical relevance of cadaveric experiments, we further evaluated the accuracy of bilateral distraction in human cadaver heads. Unified or clear clinical thresholds for MDO precision are currently lacking in the literature. Moreover, previous studies have shown varying results: Cai³⁰ reported an average angular error of 2.56° using optical navigation in sheep MDO, whereas Li⁸ documented a positional root mean square deviation (RMSD) of 0.93 mm and angular RMSDs of 4.64°, 2.03°, and 2.88° with CAD/computer-aided manufacturing (CAM) guides. In a model experiment using augmented reality (AR) navigation-assisted MDO, He et al.³¹ reported that the average angular error in the AR group was 1.94 (1.30, 2.93) with a displacement error of 1.53 ± 0.54 mm. Badiali et al³² used simulation-guided navigation to precisely control the distraction device vector in paediatric MDO, with the results showing a yaw error of 3.74° ± 3.30°, pitch error of 6.27° ± 5.32°, and median angular errors of 3.72° and 4.08°. In other studies employing craniofacial MR navigation, Shusterman et al³³ reported an average entry point deviation of 0.381 mm (XY), 0.173 mm (Z), and 0.417 mm (En) using an MR-based dynamic navigation (MR-DN) system for implant placement, with a 3D apex deviation of 0.193 mm and an angular difference of 1.852°. Tang et al³⁴ reported a mean deviation of 1.68 ± 0.92 mm between the planned and actual osteotomy planes in oral and maxillofacial tumour resection. Liu et al³⁵ reported that the mean ± SD of errors in the AR navigation system used for craniofacial fibrous dysplasia recontouring surgery was 1.442 ± 0.234 mm. The cadaveric outcomes of this study indicated the precision of MR-guided MDO.

Notably, although the SG-based approach showed accuracy comparable to that of MR in mandibular model experiments, its precision and efficiency declined in animal trials, particularly in osteotomy positioning and distractor placement. This reduction in performance is likely attributable to soft tissue interference and unstable bone-to-guide contact.³⁶ These factors may cause intraoperative guide displacement, as reflected by increased distractor centroid deviations, potentially altering the planned distraction vectors and increasing the risk of injury to dental roots and the inferior alveolar nerve. Zhu³⁷ reported that AR outperformed SGs in accuracy, reliability, and usability for specific surgical tasks such as mandibular angle osteotomy. Our findings also demonstrated enhanced reliability and precision of MR, supporting its promising clinical utility.

However, this study had several limitations. Although cadaveric specimens were included to enhance the clinical relevance of the findings, the overall sample size remained limited due to difficulties in specimen acquisition, restricting a comprehensive evaluation of real-time performance and user experience. Additionally, the assembly of the personalised positioning device may have introduced registration errors. Although the external optical-tracking system effectively addressed marker occlusion and tracking instability in complex intraoral environments, it also increased hardware complexity, which may be a limitation of the current system. Furthermore, intra-observer and inter-observer reproducibility were not evaluated in this study. Therefore, future studies should focus on enhancing the precision of 3D-printed components to minimise assembly errors, expanding the sample size for more robust validation, and refining optical-tracking technology to improve portability. Martin-Gomez³⁸ demonstrated that the built-in sensors of HoloLens 2 allow real-time surgical tool tracking without external hardware, highlighting the potential for future autonomous OST-HMD–based navigation systems. Further advances in virtual–real fusion, spatial perception, and artificial intelligence (AI)-assisted visualisation may also help overcome challenges such as virtual obstruction and depth perception limitations,³⁹ facilitating the broader application of MR navigation in craniomaxillofacial and other precision surgeries.

Conclusion

In this comparative preclinical study, the MR navigation system showed fewer errors than the FH approach in distractor placement, distraction vector, and osteotomy, and it outperformed the SG-based approach in distractor positioning and osteotomy accuracy. In comparison with TN, the MR system excelled in distractor positioning and osteotomy accuracy and reduced the operative time. The MR system also enabled accurate intraoperative execution of preoperative distraction vectors, indicating that it is a feasible and promising approach for clinical translation of MR navigation technology in MDO. However, further clinical studies are needed to validate these preclinical outcomes.

Consent for publication

Not applicable.

Authors' information

Not applicable.

Author contributions

Conceptualisation, methodology, investigation, data curation, formal analysis, writing—original draft, writing—review and editing: Ma, He, and Liu. Data curation, formal analysis, writing—original draft: Shang. Resources, supervision, literature review, writing—review and editing: Gong. Conceptualisation, supervision, funding acquisition, writing—review and editing: Huang.

Ethics statement

This study was conducted in accordance with the Declaration of Helsinki and the ethical standards for animal and cadaveric research. All animal experiments were approved by the Laboratory Animal Ethics Committee of Guangxi Medical University in accordance with the Guideline for Ethical Review of Laboratory Animal Welfare (GB/T 35892-2018). Cadaveric specimens were obtained from the Department of Anatomy, Guangxi Medical University, with approval from the Ethics Committee of Guangxi Medical University Hospital of Stomatology (approval no. 2023080). This study was registered with the Chinese Clinical Trial Registry on March 14, 2025 (registration no. ChiCTR2500098863).

Declaration of generative AI and AI-assisted technologies in the writing process

This article was written by the authors independently without using any AI tools or software to generate, edit, or modify the content.

Conflict of interest

The authors declare that they have no conflict of interest.