Mixed-reality, heads-up navigation for percutaneous pedicle screw placement: a prospective accuracy and workflow study

Abstract

Background

Mixed-reality (MR) head-mounted navigation projects a three-dimensional hologram of the patient’s spine directly into the surgeon’s field of view, potentially enhancing the precision and speed of percutaneous pedicle screw fixation. High-quality prospective data comparing MR with conventional fluoroscopy are scarce, prompting the present randomized study.

Results

Sixty-one adults (274 screws) were allocated to MR (30 patients, 136 screws) or fluoroscopy (31 patients, 138 screws). MR navigation yielded a greater proportion of accurately placed screws (96.3% vs. 89.9%; absolute difference 6.4%; p = 0.04) and shortened mean operative time by 28 min (83 ± 11 min vs. 111 ± 14 min; p < 0.001). Entry-point, angular and depth deviations were all significantly smaller with MR, while intra-operative fluoroscopic images fell by 42%. Early complications occurred in 3.3% of MR cases and 16.1% of controls, and holographic pre-operative counselling improved patient understanding and satisfaction (p < 0.001).

Conclusions

MR head-mounted navigation is feasible, safe and more precise than fluoroscopic guidance, delivering faster procedures, fewer fluoroscopic shots and a short learning curve. These advantages support MR as a cost-effective next-generation guidance modality for minimally invasive spine surgery.

Introduction

Percutaneous pedicle screw fixation (PPSF) is a minimally invasive spinal fixation technique widely utilized in spinal trauma, degenerative diseases, and deformity correction [1, 2]. This technique involves the placement of pedicle screws through small incisions under fluoroscopic guidance, avoiding the paraspinal muscle dissection required in traditional open surgery, thereby reducing surgical trauma and postoperative complications [3]. However, this procedure demands high technical proficiency and extensive clinical experience, with a steep learning curve [3]. Improper technique may result in pedicle breach, leading to neurological injury or fixation failure [4, 5].

In recent years, computer-assisted navigation and robotic-assisted surgery have played important roles in improving screw placement accuracy, but these technologies still possess certain technical limitations [6].These include high dependency on equipment and operative environment, steep learning curves, visual field interruption, and surgeon attention diversion [7, 8].

Mixed reality (MR) technology represents a novel digital holographic imaging technique that emerged following virtual reality (VR) and augmented reality (AR) technologies. MR can integrate information from the real world with virtual images, displaying virtual information within real-world scenarios to create new visualization environments [9, 10]. Holographic imaging technologies represented by MR have found applications across multiple medical fields, encompassing medical education, surgical skills training, preoperative planning, and intraoperative precision guidance, demonstrating promising potential for personalized treatment in minimally invasive spinal techniques [11].

In percutaneous pedicle screw placement procedures, MR technology can utilize real-time imaging combined with patient CT or MRI data to create three-dimensional models, providing intuitive intraoperative navigation guidance. Through MR head-mounted displays (HMDs), surgeons can view the real-time relationship between screw position and patient anatomy without requiring visual field transitions or dependence on external monitors. Currently, China lacks research on MR technology-assisted PPSF implementation. The primary objective of this study is to evaluate the safety and feasibility of MR technology-assisted PPSF.

Method

This study is a prospective, randomized controlled trial (RCT) designed to evaluate the efficacy and safety of MR-assisted percutaneous pedicle screw placement technique. The study obtained institutional review board approval from the participating center, and all participants provided informed consent prior to enrollment. The primary objectives were to assess clinical accuracy of screw placement, surgical efficiency, and complication rates, while secondary outcomes included preoperative patient understanding of the procedure.

Randomisation and allocation

Inclusion criteria comprised patients under 60 years of age scheduled for single-segment spinal fusion or percutaneous pedicle screw placement involving no more than three consecutive levels (L1-S1). Exclusion criteria were severe osteoporosis (a T score of less than − 3.5), other bone abnormalities (e.g., osteogenesis imperfecta, tumors, infection, or malignancy), spondylolisthesis grade III or IV, neurologic deficiency, and individuals who were unwilling to sign written informed consent and did not agree to participate in the study. A total of 61 patients ultimately provided informed consent and were enrolled in the study. All 61 patients were aged between 25 and 60 years(Table 1). Given the modest sample size, subjects were stratified and randomized based on the number of implanted screws (surgical segments) and bone density to ensure balanced and comparable baseline characteristics between groups. The experimental group underwent MR technology-assisted PPSF, while the control group received conventional pedicle screw placement procedures. Although no universally accepted threshold has been established, an alignment error of less than 2 mm was predefined as clinically acceptable in the present study, based on previously reported registration and navigation accuracy in augmented-reality and navigation-assisted spinal instrumentation workflows.

Table 1.

Baseline Characteristics presents the baseline characteristics comparison between the experimental group (MR group) and the control group (CG group), including age, sex, BMI, and bone mineral density. No significant differences were found between the two groups (P > 0.05), confirming the balance of randomization

	EG (n = 30)	CG (n = 31)	P-Value
Age	48.6 ± 8.2	49.4 ± 8.7	0.71
Sex	16	14	0.52
BMI	24.8 ± 3.2	25.2 ± 2.6	0.59
BMD T-score<2.5	6	6	-
Screws inserted ≤ 4	22	24	-
Smoking	5	6	0.78
Diabetes	3	4	1.00

All surgeries were performed by the same group of spine surgeons, each with over 10 years of clinical experience in spinal surgery and holding senior professional titles (attending or associate professor level and above). Prior to the procedures in the experimental group, all surgeons underwent theoretical training and hands-on practice using the mixed reality (MR) navigation system on simulation models to ensure procedural standardization and consistency. For the control group, the same surgeons performed pedicle screw placement using conventional fluoroscopy-guided (C-arm) techniques, without prior MR navigation training or simulation practice. This approach minimized inter-surgeon variability and ensured comparability between groups. Physical three-dimensional printed models were generated only for preoperative anatomical review and surgical rehearsal and were not used for intraoperative navigation or quantitative outcome assessment.

Mixed-reality navigation workflow

The overall MR-assisted pedicle screw navigation workflow consisted of five sequential stages: preoperative imaging acquisition and three-dimensional reconstruction, anatomical feature-point planning, holographic model registration, intraoperative MR-guided screw placement, and postoperative accuracy evaluation.

First, patient-specific spinal three-dimensional models were reconstructed from preoperative DICOM imaging datasets using segmentation and surface-reconstruction algorithms. Anatomical feature points recognizable on both the reconstructed model and the patient anatomy were predefined during preoperative planning(Fig. 1).

Fig. 1.

Preoperative surgical planning using 3d models preoperative planning is conducted using Mimics Medical software, which creates a 3D spinal model from the patient’s CT and MRI data. This model provides decision support for the optimal trajectory, length, and diameter of the pedicle screws

Second, reconstructed spinal models were imported into the mixed-reality navigation platform and visualized intraoperatively through the HoloLens 2 head-mounted display. A custom model-positioning device was used to facilitate holographic registration between the virtual spinal model and the patient anatomy(Fig. 2).

Fig. 2.

Preoperative Explanation Using the System The HoloLens 2 head-mounted mixed reality device is used for preoperative explanation, helping patients better understand the surgical process. By integrating the 3D virtual model with the patient’s actual anatomy, the device enhances the patient’s comprehension and trust in the surgery

Third, the surgeon sequentially aligned virtual fiducial points with predefined anatomical landmarks on the patient body surface according to the planned registration order. Following feature-point confirmation, spatial coordinate transformation was automatically performed using a proprietary registration algorithm to achieve holographic alignment between the virtual model and the patient anatomy(Figs. 3 and 4).

Fig. 3.

Intraoperative mixed-reality navigation for percutaneous pedicle screw placement. The mixed-reality system projects the reconstructed three-dimensional spinal model and planned screw trajectories into the surgeon’s field of view through the head-mounted display. The holographic model is registered with the patient anatomy, allowing real-time visualization of the spatial relationship between the pedicle, vertebral structures, and planned screw paths during implantation. This heads-up navigation workflow provides intuitive trajectory guidance and supports accurate percutaneous pedicle screw insertion while reducing reliance on repeated fluoroscopic confirmation

Fig. 4.

Intraoperative Direct Operation with MR System The surgeon wears the MR device during surgery, which provides real-time visualization of the patient’s anatomical structures and virtual models. This enables the precise insertion of pedicle screws directly in the surgeon’s line of sight, eliminating visual interference and improving operational accuracy

Fourth, pedicle screw placement was performed under MR-guided holographic visualization. Intraoperative fluoroscopy was additionally used for registration verification and trajectory confirmation when necessary.

Finally, postoperative CT imaging was obtained for quantitative assessment of screw placement accuracy, including pedicle breach grading and trajectory deviation analysis.

Radiographic analysis

Preoperative CT imaging was used for three-dimensional model generation and surgical planning. Intraoperative fluoroscopy was primarily used to verify holographic registration accuracy and screw trajectory consistency during implantation. Postoperative CT imaging was subsequently obtained for formal quantitative evaluation of pedicle screw placement accuracy. Screw accuracy was assessed using the Gertzbein-Robbins classification system. Entry-point deviation was defined as the three-dimensional Euclidean distance between the planned pedicle screw entry point on the preoperative holographic model and the actual postoperative screw entry point identified on postoperative CT reconstruction. Angular deviation was defined as the angular difference between planned and actual screw trajectories in three-dimensional space. Depth deviation represented the linear difference between planned and actual screw insertion depth. All radiographic measurements were independently evaluated by two experienced spine surgeons blinded to group allocation [12].

Outcome

Primary outcome measures included pedicle screw placement accuracy, operative efficiency, and perioperative complications. Procedural efficiency was evaluated using implantation time per screw and total intraoperative surgical duration. Operative time measurements reflected intraoperative surgical duration only and did not include preoperative segmentation, holographic model generation, MR system preparation, or registration time. Fluoroscopy frequency was recorded as a surrogate indicator of radiation exposure. Patient satisfaction was assessed using an institutionally developed postoperative questionnaire evaluating subjective understanding of the surgical procedure, perceived technological acceptance, and overall perioperative experience(Fig. 5). Questionnaires were completed prior to hospital discharge(postoperative day 3). The questionnaire was designed for exploratory assessment purposes and was not formally externally validated. Learning-curve analysis was performed using sequential case-based regression analysis of operative time and screw-placement accuracy across MR-assisted procedures. Analyses were conducted using pooled surgeon data because all participating surgeons possessed comparable senior-level minimally invasive spine surgery experience.Learning-curve analysis was performed using sequential case-based nonlinear regression analysis of operative time and screw-placement accuracy across MR-assisted procedures. Trend stabilization was defined as the point at which consecutive procedural improvements reached a plateau with minimal incremental reduction in operative duration.

Fig. 5.

Patient Satisfaction with MR Preoperative Education A chart depicting patient satisfaction scores following preoperative education using the MR system.The MR group demonstrated significantly improved procedural understanding, perceived confidence, and overall satisfaction compared with the control group

Statistical analysis

Continuous variables were expressed as mean ± standard deviation and compared using Student’s ttest. Categorical variables were analyzed using the chi-square test or Fisher’s exact test as appropriate. Statistical significance was defined as P < 0.05. Statistical analyses were performed using SAS software version 9.14. The sample size was primarily determined based on expected differences in pedicle screw placement accuracy and operative efficiency. The study was not specifically powered to detect statistically significant differences in complication rates.

Result

Baseline characteristics

Sixty-one participants were randomized (MR = 30, Control = 31). No significant between-group differences were observed in age, sex ratio, BMI, osteoporosis rate, smoking status, or diabetes (all P > 0.50), confirming adequate randomization(Table 1).

Primary outcomes

A total of 274 pedicle screws were analyzed (MR = 136, Control = 138). MR navigation achieved 96.3% Grade A + B accuracy compared to 89.9% in controls (risk difference + 6.4%, P = 0.04) (Table 2). Mean operative time was 83 ± 11 min with MR versus 111 ± 14 min with conventional fluoroscopy (Δ − 28 min, P < 0.001) (Table 3). Quantitative three-dimensional analysis demonstrated significantly smaller entry-point, angular, and depth deviations in the MR group (all P < 0.001) (Table 4).

Table 2.

Pedicle Screw Accuracy Scores shows the postoperative pedicle screw accuracy scores. The MR group achieved significantly higher accuracy (Grade A and B) compared to the control group (CG group) (P = 0.04). There were also significant differences in entry point, angular, and depth deviations between the groups

	EG(n = 30)	CG (n = 31)	P-Value
Segment	22(73%)	24(77%)	0.71
Screw insertion success rate (%)	96%	90%	0.04*
Grades
A	81	62	0.03*
B	50	62
C	5	11
D	0	3
C-arm fluoroscopy count (n)	3.1 ± 0.7	11.5 ± 1.7	<0.001*

P<0.05*

Table 3.

Comparison of Surgical Duration between the MR group and the control group for cases with different screw numbers. The MR group had significantly shorter average surgical times, indicating that the use of MR technology can reduce the overall procedure duration

	Number of screws	Cases	Mean ± SD (min)
Experimental	4 screws	22	78 ± 10
Experimental	6 screws	8	98 ± 12
Total Experimental		30	83 ± 11
Control	4 screws	24	105 ± 13
Control	6 screws	7	128 ± 15
Total Control		31	111 ± 14

Table 4.

Accuracy and postoperative complication Incidence compares the accuracy of pedicle screw placement and the incidence of postoperative complications between the MR group (EG) and the control group (CG). The MR group showed significantly better accuracy with smaller entry point, angular, and depth deviations, and had a lower complication rate (3.3%) compared to the control group (16.1%)

Complication Type	EG (n = 30)	CG (n = 31	P -Value
Neurological Injury	0	1	1.00
Vascular Injury	1	2	1.00
Infection	0	1	1.00
Reoperation Required	0	1	1.00
Total Complications	1 (3.3%)	5 (16.1%)	0.19
Parameter Type
Entry Point Deviation (mm)	1.4 ± 0.5	2.9 ± 0.7	< 0.01*
Angular Deviation (°)	2.8 ± 0.9	4.6 ± 1.2	< 0.01*
Depth Deviation (mm)	1.6 ± 0.6	3.2 ± 1.0	< 0.01*

P<0.05*

Safety

Overall complication incidence was 3.3% (1/30) with MR versus 16.1% (5/31) in controls; however, this difference was not statistically significant due to limited statistical power (P = 0.19) (Table 4).

Patient surgical satisfaction

Preoperative patient education using the MR system improved preoperative understanding of the procedure, anxiety relief, and overall satisfaction, while pain relief was comparable (all P < 0.001 except pain, P = 0.83) (Table 5).

Table 5.

Patient satisfaction comparison compares patient satisfaction scores between the MR group (EG) and the control group (CG). The MR group had significantly higher scores in preoperative understanding, anxiety relief, and overall satisfaction with the surgical outcome, leading to a higher total satisfaction score

Evaluation Item	EG (n = 30)	CG (n = 31)	P -Value
Understanding of surgical procedure before surgery	4.6 ± 0.6	3.9 ± 0.7	< 0.001*
Relief of anxiety before surgery	4.4 ± 0.7	3.5 ± 0.8	< 0.001*
Pain relief after surgery	3.9 ± 0.8	3.8 ± 0.8	0.832
Overall satisfaction with surgical outcome	4.4 ± 0.5	3.6 ± 0.7	< 0.001*
Total score (out of 20)	17.3 ± 1.2	14.9 ± 1.6	< 0.001*

P<.05^*

Subgroup analysis and learning curve assessment

Subgroup analysis demonstrated that the time-saving effect persisted in both four-screw and six-screw cases (interaction P = 0.27), suggesting robustness across surgical complexity levels. Learning-curve regression revealed a plateau after case 7 (Fig. 5), indicating rapid proficiency acquisition.

Discussion

Principal findings

In this prospective randomised trial, mixed-reality (MR) head-mounted navigation achieved a 96.3% Grade A + B screw-placement accuracy versus 89.9% with conventional fluoroscopy, and shortened mean operative time by 28 min while reducing fluoroscopic shots by 42%. These gains were consistent across four- and six-screw subgroups and were accompanied by a rapid learning-curve plateau after the seventh MR case, indicating quick adoption by experienced surgeons. Although early complications were numerically lower with MR (3.3% vs. 16.1%), the study was under-powered to confirm significance. Strengths of the trial include a randomised design, CT-based 3-D accuracy assessment and concurrent evaluation of patient-reported outcomes. Limitations comprise the single-centre setting, modest sample size and absence of direct radiation-dose measurement, which warrant cautious generalisation.

Although MR-assisted pedicle screw placement demonstrated improved procedural accuracy and operative efficiency in the present study, several important technical limitations should be acknowledged. The current MR workflow depends on stable optical tracking, gesture-recognition accuracy, and reliable fiducial feature-point identification. Environmental lighting conditions, temporary tracking instability, obstruction of the visual field, or inaccurate feature-point selection may potentially reduce holographic registration accuracy. Furthermore, because the holographic model was generated from preoperative CT imaging acquired in a static position, anatomical discrepancies may occur intraoperatively following prone positioning or soft tissue deformation. To minimize these effects, repeated fluoroscopic verification of holographic alignment was performed during surgery, and minor positional adjustments of the hologram could be achieved through synchronized tablet-based controls. Nevertheless, dynamic anatomical changes remain a potential source of navigation error. In addition, although MR-assisted screw placement reduced intraoperative implantation time, the complete workflow additionally required preoperative image segmentation, holographic model generation, feature-point planning, and hologram registration. These preparatory procedures were not included in the operative-time analysis and should be considered when evaluating total workflow efficiency. Direct radiation-dose measurements for surgeons and patients were not routinely available from the fluoroscopic system used during the study period. Consequently, fluoroscopy frequency was used as a surrogate indicator of radiation exposure. This indirect assessment may not fully reflect the true radiation reduction achieved using MR navigation.

Comparison with the literature

Compared with conventional optical navigation systems such as BrainLab and robotic-assisted spinal navigation platforms, the MR workflow used in the present study provided direct heads-up visualization without requiring repeated attention shifts toward external navigation monitors [12, 13]. Unlike robotic-assisted systems requiring dedicated robotic-arm integration and complex workstation interaction, the current MR platform utilized intuitive gesture-based interaction and surgeon-controlled holographic visualization. This workflow may partially explain the relatively short learning curve observed in the present study [14]. Previous robotic-assisted spine surgery studies have reported learning-curve stabilization after approximately 10–20 cases(Fig. 6). In comparison, the current MR-assisted workflow demonstrated apparent stabilization after approximately seven cases. However, this finding should be interpreted cautiously because participating surgeons were already highly experienced in minimally invasive spinal procedures.Because participating surgeons already possessed extensive minimally invasive spinal surgery experience, the observed learning curve likely reflects adaptation to the MR interface rather than acquisition of fundamental pedicle screw instrumentation skills [15]. Despite these advantages, conventional navigation platforms currently provide more mature automatic registration algorithms and validated multi-centre workflow integration. Additional studies are therefore necessary to determine whether MR navigation can achieve comparable reproducibility across broader surgical environments.

Fig. 6.

Surgeons’ Learning Curve with MR Technology A graph illustrating the learning curve for surgeons using MR navigation. Operative performance demonstrated relative stabilization after approximately seven cases

Clinical implications and future directions

The present study was conducted at a single institution using a standardized workflow and experienced spine surgeons familiar with minimally invasive spinal procedures. In addition, the workflow required dedicated preoperative model processing and familiarity with the Jinshu holographic platform, which may limit immediate reproducibility across centres lacking similar technical infrastructure or engineering support. Furthermore, the study excluded patients with severe osteoporosis, major spinal deformity, and highgrade spondylolisthesis. Therefore, applicability of MR navigation to more complex spinal pathology and broader patient populations remains to be established. The current workflow relied on the Microsoft HoloLens 2 platform, which is expected to lose long-term commercial support in coming years. Future studies should therefore evaluate cross-platform compatibility using emerging MR/AR hardware systems to ensure broader clinical applicability and longterm sustainability [16, 17].

Conclusion

Mixed-reality heads-up navigation demonstrated improved pedicle screw placement accuracy, reduced intraoperative implantation time, and decreased fluoroscopy frequency compared with conventional fluoroscopy-assisted techniques. The present workflow provides an intuitive surgeon-controlled visualization platform that may enhance intraoperative spatial perception during minimally invasive spinal instrumentation. However, additional multi-centre studies with larger patient populations are necessary to validate reproducibility, evaluate long-term clinical outcomes, and further optimize registration robustness and workflow integration. MR-assisted spinal navigation represents a promising next-generation image-guidance modality for minimally invasive spine surgery.

Author contributions

Ruili Zhuo: Conceptualisation, surgery, data interpretation, manuscript drafting Hongrui Liu: data acquisition Wei Liang: Patient recruitment, critical revision Chen Xu: Image processing, statistical analysis Shaodong Sun: Hardware integration, figure preparation Guoyu li: software validation All authors read and approved the final manuscript.

Funding

This work was supported by the Henan Provincial Science & Technology Project (Grant No. 242102310489). The funder had no role in study design; in the collection, analysis or interpretation of data; in the writing of the manuscript; or in the decision to submit the article for publication.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Puyang Oilfield General Hospital. Written informed consent was obtained from all participants prior to enrolment.

Consent for publication

All participants provided explicit consent for the publication of anonymised clinical information and imaging data.

Competing interests

The authors declare no competing interests.