Preliminary evaluation of mixed reality navigation with single-vertebra registration in spinal surgery

Abstract

Objective

To develop a mixed reality navigation system based on single-vertebra registration and to preliminarily validate its feasibility in clinical surgery, providing a potential reference for spinal surgical navigation registration methods.

Methods

We developed a portable, stand-alone, optical-based mixed reality navigation system for spinal surgery using a head-mounted display. The system integrates registration and real-time localization functions, enabling single-vertebra registration via anatomical landmarks and interactive visualization through the HoloLens 2 platform. System performance was evaluated in both laboratory and clinical settings by measuring localization time, fiducial registration error (FRE), and target registration error (TRE). Preoperative assessments were conducted on patient-specific 3D-printed vertebral models, and intraoperative evaluations were performed on exposed vertebrae during surgery. Additional trials with novice surgeons were performed to assess the learning curve. For selected cases requiring pedicle screw placement, surgeons qualitatively evaluated the intraoperative visualization of preplanned screw trajectories in terms of anatomical congruence, feasibility, and display clarity.

Results

A total of 37 spinal surgery patients were included, covering cervical, thoracic, and lumbar segments. The average preoperative navigation time based on 3D-printed models was 2.96 ± 0.26 min, compared with 3.11 ± 0.25 min intraoperatively (p = 0.002). The FRE was 1.37 ± 0.54 mm preoperatively and 1.42 ± 0.52 mm intraoperatively (p = 0.178). The TRE was 1.52 ± 0.56 mm preoperatively and 1.75 ± 0.61 mm intraoperatively (p = 0.001). Among four pedicle screw placement cases, three were rated satisfactory intraoperatively, and one showed an approximately 5 mm deviation due to minor patient positional changes. The learning curve indicated that procedure time stabilized at approximately 2.7 min after about 15 practice cases.

Conclusion

The single-vertebra registration-based mixed reality navigation system enables rapid and accurate vertebral localization in both preoperative and intraoperative settings, with TREs within clinically acceptable limits. The system is portable, easy to operate, and provides an intuitive, immersive navigation interface without relying on external tracking devices. These features suggest its potential as a cost-effective and versatile tool for spinal surgery navigation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40001-025-03307-7.

Introduction

With the advancement of image-guided techniques, surgical navigation has significantly improved the safety and precision of cranial, spinal, and other orthopedic procedures [1–3]. By using preoperative CT or MRI scans, this technology accurately depicts the anatomical structures of the target region, assisting surgeons in precisely localizing and identifying critical anatomy during surgery, and reducing injury to important adjacent tissues [4, 5].

Currently, optical navigation systems are the most widely used in clinical practice and can provide high spatial accuracy [6]. These systems track the real-time position of surgical instruments (e.g., a probe) using optical tracking devices and register them to preoperative images. Point-to-point or surface-to-surface rigid registration is commonly used to map the entire preoperative dataset to the patient’s intraoperative position [7]. The instrument position is then displayed on an external monitor to guide the surgeon. However, this approach has several limitations. First, optical navigation equipment is bulky and expensive, limiting adoption in small and medium-sized hospitals. Second, it treats the preoperative image as a static whole for rigid transformation. In spinal surgery, changes in patient posture or the natural mobility of the spine can create spatial discrepancies between preoperative images and the actual intraoperative anatomy, reducing registration accuracy [8]. Finally, surgeons must shift their gaze from the operative field to a remote display, causing “hand–eye separation” and reducing intuitiveness and efficiency[9].

In recent years, the rapid development of mixed reality (MR) technology has introduced a new solution for surgical navigation [10, 11]. Using a head-mounted display (HMD), MR overlays preoperatively constructed three-dimensional (3D) models onto the real anatomy during surgery, providing an immersive and intuitive navigation experience. In previous study, we evaluated the accuracy of MR-based navigation in neurosurgery, demonstrating that it met the clinical precision requirements for most neurosurgical procedures [12]. In spinal procedures, MR offers strong interactivity and flexibility, showing great potential in assisting intraoperative anatomical identification, pathway planning, and implant guidance. However, most systems remain in laboratory or simulation testing, with limited evidence for stable, real-time performance in actual surgeries [13, 14].

To address this gap, we developed a spinal mixed reality (SMR) navigation system and proposed a rapid, “on-demand single-vertebra registration” strategy. This method creates a high-precision 3D model of the target vertebra preoperatively and performs quick intraoperative registration using exposed bony landmarks. The system leverages the built-in simultaneous localization and mapping (SLAM) capability of the Microsoft HoloLens 2 to sense intraoperative environmental changes, improving robustness and adaptability to spinal position shifts [15]. The navigation process does not rely on traditional optical tracking or invasive markers, offering lightweight hardware, flexible deployment, and greater suitability for clinical settings.

To our knowledge, few studies have reported MR systems capable of achieving stable registration and real-time navigation in live spinal surgery. This study aimed to conduct a preliminary clinical evaluation of the SMR system, assessing its feasibility in localizing spinal anatomy intraoperatively, and to lay the groundwork for future applications in instrument navigation and pedicle screw placement.

Materials and methods

System architecture

The SMR system consists of three components: the HoloLens 2 hardware, navigation software, and a spinal probe tool. The navigation software was developed on the Unity platform (Unity Software, Inc.) and features a user interface that supports model import, model property adjustment (e.g., transparency, color), registration, and target point setting. The spinal probe tool comprises a probe holder, a probe tip (1 mm in diameter), and four optical reflective spheres. The probe holder was 3D-printed to mount the reflective spheres in a non-symmetrical arrangement, with the distance between any two spheres exceeding 2 cm. When the SMR system is activated, it automatically engages the device’s time-of-flight (TOF) camera to scan the physical space in real time, detect the position of the probe tool, and align the preconfigured virtual probe model with the recognized tool position [16] (Fig. 1). The software also includes a built-in voice control function, allowing the surgeon to place markers and perform registration through predefined voice commands during use.

Fig. 1.

SMR hardware components. A Appearance of the HoloLens 2; B 3D-printed surgical tracking tool designed for spinal procedures; C virtual model of the tracking tool in the SMR environment; D when multiple tools appear simultaneously within the HoloLens 2 field of view, the SMR system identifies only the designated tool, while non-target tools (yellow arrow) are ignored

Patient selection

Patients who underwent spinal surgery at our institution between January and June 2025 were considered for inclusion. This study was approved by the local Ethics Committee.

Inclusion criteria

• Age ≥ 18 years.

• Preoperative imaging of sufficient quality to clearly and accurately depict vertebrae and major neurovascular structures.

• Patients and their families were fully informed of the surgical plan and study protocol, consented to the procedure, and signed a written informed consent form.

Exclusion criteria

• Severe medical conditions that precluded surgery, such as cardiovascular disease or coagulation disorders.

• Pregnancy or lactation.

• Refusal of the patient or family to participate in the study.

Data processing

All patients underwent preoperative spiral CT scanning (Ingenuity Core; Philips Healthcare) with the following parameters: a slice thickness of 1 mm, a reconstruction interval of 0 mm to avoid overlap, and a field of view (FOV) adjusted according to patient size, typically set at 250 mm × 250 mm. The pixel matrix was set to 512 × 512 to ensure high resolution. For window settings, a bone window of 3000/400 HU and a soft tissue window of 400/40 HU were used. The scan duration was approximately 10–15 s.

The acquired images were saved in DICOM format, anonymized, and imported into 3D Slicer (https://www.slicer.org, version 5.8) for 3D reconstruction and anatomical landmark annotation [17]. The workflow was as follows: Based on the affected spinal segment, a high-precision 3D reconstruction of the target vertebra was performed. Five distinct anatomical landmarks (e.g., spinous process, left and right transverse processes) were manually selected for each vertebra and precisely annotated in 3D Slicer. Four of these points were used for intraoperative spatial registration between the virtual model and the real anatomy, while the remaining point served for intraoperative localization accuracy assessment. After reconstruction, the 3D model was exported in OBJ format. The model was used for two purposes: (1) fabrication of a physical 3D-printed model and (2) import into the SMR system for model visualization and intraoperative guidance in the mixed reality environment. Figure 2 illustrates the complete data acquisition and localization workflow.

Fig. 2.

Workflow of the study based on the SMR system, including preoperative CT acquisition, 3D reconstruction and landmark annotation, model export, 3D printing, and intraoperativeMR navigation

Accuracy evaluation

Preoperative evaluation

The patient-specific 3D-printed vertebral model was first secured to an operating platform. The SMR system was then activated, and the surgeon used the dedicated spinal probe tool to sequentially select four predefined anatomical landmarks on the model surface (e.g., spinous process, transverse processes). The procedure was as follows: the surgeon positioned the probe tip on the target point and triggered the system to record its spatial coordinates using the voice command “Add.” A virtual point was then generated at the probe tip location. This process was repeated until all four points were added. The surgeon then issued the voice command “Register,” prompting the system to automatically perform rigid registration between the virtual vertebral model and the physical model, and output the fiducial registration error (FRE). FRE was defined as the root mean square (RMS) of the Euclidean distances between the surgeon-selected points and their corresponding predefined points in the model, reflecting the spatial alignment accuracy between the virtual and physical models. After registration, the HoloLens 2’s built-in SLAM feature anchored the virtual vertebral model stably in the physical space. Under SMR guidance, the surgeon then used the probe to locate the remaining landmark, and the Euclidean distance between the probe tip and the true point was recorded as the target registration error (TRE). Each vertebra underwent three independent localization trials, and the mean TRE was used for analysis. Localization time was defined as the total duration from placing the first landmark to completing the target point localization. To further evaluate the learning efficiency of users with different levels of experience, we additionally recruited three surgeons with no prior experience using MR navigation (novice group). Following the exact same protocol as the experienced surgeons, each novice repeatedly performed independent registration and localization trials on the 3D-printed model. FRE, TRE, and localization time were recorded for each trial, and performance trends over successive trials were analyzed to plot and compare learning curves. Video 1 demonstrates the process of using the SMR system from a first-person perspective.

Intraoperative evaluation

Once the target vertebra was fully exposed during surgery, the operating surgeon performed registration and localization following the same procedure as in the preoperative evaluation. The probe tool used intraoperatively was sterilized with low-temperature plasma to meet aseptic requirements. All FRE, TRE, and operation times during localization were recorded and documented by an independent observer.

Qualitative evaluation

For patients requiring pedicle screw placement, the surgeon defined the screw entry point and insertion trajectory during 3D modeling, creating a virtual channel model. After intraoperative registration, the surgeon qualitatively assessed the virtual path in three dimensions:

• Anatomical congruence—the degree to which the virtual channel aligned with exposed bony landmarks (e.g., transverse process, spinous process).

• Instrumentation feasibility—whether the planned path allowed for convenient insertion of standard surgical instruments intraoperatively.

• Visual clarity of the path—the stability and intuitiveness of the mixed reality display.

It should be noted that in this study, the final surgical trajectory and procedure were determined by the surgeon’s clinical judgment. The SMR system output served only as an intraoperative adjunct.

Sample size estimation

The sample size was determined based on the actual surgical volume of the medical center and preliminary test results. Using G*Power software, the significance level was set at α = 0.05 and statistical power at 0.90. The expected mean difference in TRE between preoperative and intraoperative measurements was 0.30 mm, with a standard deviation of 0.50 mm (Cohen’s d ≈ 0.60). According to the paired t-test sample size calculation, the minimum required sample size was 32 cases. Considering an anticipated 5% loss to follow-up or missing data, the final minimum enrollment target was set at 34 cases.

Statistical analysis

All data were analyzed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA). Continuous variables were tested for normality (Shapiro–Wilk test). Normally distributed data are presented as mean ± SD and compared using paired t-tests, while non-normally distributed data are shown as median (IQR) and compared using the Wilcoxon signed-rank test. Agreement between preoperative and intraoperative localization was assessed using Bland–Altman analysis. For inter-doctor comparisons, the Friedman test with post hoc Wilcoxon signed-rank tests (Bonferroni correction) was applied.

Results

A total of 37 cases were included in this study, including 20 males and 17 females. The lesions were located in the cervical, thoracic, and lumbar spine. A qualitative observation of pedicle screw implantation was conducted in 4 patients. The demographic and surgical information is summarized in Table 1.

Table 1.

Summary of patient demographic and surgical information

Variable	Statistic
Total patients	37
Gender (male/female)	20/17
Age
Range (years)	29–72
Mean ± SD (years)	51 ± 12
Lesion location (cases)
Cervical spine	8
Thoracic spine	16
Lumbar spine	13
Lesion type (cases)
Herniated disc	11
Spinal stenosis	6
Spinal fracture	5
Spinal tumor	15
Screw evaluation (cases)	4
Surgical position	Prone position

At least four anatomical landmarks could be identified on vertebrae from different segments (Fig. 3A–F). Using point-based registration, single-vertebra registration was successfully performed both preoperatively and intraoperatively (Fig. 3G–L).

Fig. 3.

Vertebral reconstruction and 3D printing based on patients’ preoperative CT scans. A–C Three-dimensional reconstructions of cervical, thoracic, and lumbar vertebrae showing anatomical landmarks used for navigation registration. D–F Corresponding 3D-printed models of the cervical, thoracic, and lumbar vertebrae. G View from the SMR system showing the imported virtual model before registration; the virtual and physical models are independent at this stage. H Registration completed, with the virtual and physical models aligned and overlaid in physical space. I Target identification and localization using a surgical probe tool, with real-time display of the distance from the probe tip to the target (white arrow). Intraoperative views: surgeon wearing the HoloLens headset (J); overlay of the virtual model onto the real vertebra (K); adjusting model transparency to observe pedicle screw placement (L)

Preoperative and intraoperative validations were completed using the SMR system. The results are summarized in Table 2.

Table 2.

Results from 3D-printed and intraoperative vertebrae tests

Variable	Preoperative	Intraoperative	Statistic	P-value
Time (min)	2.96 ± 0.26	3.11 ± 0.25	t = −3.3	0.002
FRE (mm)	1.37 ± 0.54	1.42 ± 0.52	t = −1.4	0.178
TRE (mm)	1.52 ± 0.56	1.75 ± 0.61	t = −3.9	0.001

Preoperative localization time based on 3D-printed vertebrae was 2.96 ± 0.26 min (95% CI 2.88–3.05; range 2.50–3.60 min). Intraoperative localization time was 3.11 ± 0.24 min (95% CI 3.03–3.20; range 2.60–3.70 min). Paired t-test showed a significant difference (t = −3.3, p = 0.002), indicating that intraoperative localization took longer than preoperative localization for the same target (Fig. 4A).

Fig. 4.

Comparison of preoperative and intraoperative navigation localization metrics. A Scatter plot with connecting lines showing localization time before and during surgery. B Scatter plot with connecting lines showing FRE before and during surgery. C Violin plot of preoperative and intraoperative FRE distributions. D Scatter plot with connecting lines showing TRE before and during surgery. E Violin plot of preoperative and intraoperative TRE distributions. F Bland–Altman plot comparing preoperative and intraoperative TRE, showing mean bias and 95% limits of agreement

Preoperative FRE based on the printed model was 1.37 ± 0.54 mm (95% CI 1.19–1.55; range 0.30–2.30 mm). Intraoperative FRE based on patient vertebrae was 1.42 ± 0.52 mm (95% CI 1.25–1.59; range 0.40–2.30 mm). The paired t-test showed no significant difference (t = −1.4, p = 0.178), indicating comparable registration accuracy pre- and intraoperatively (Fig. 4B–C).

Preoperative TRE based on the printed model was 1.52 ± 0.56 mm (95% CI 1.33–1.70; range 0.45–2.50 mm). Intraoperative TRE based on patient vertebrae was 1.75 ± 0.61 mm (95% CI 1.55–1.96; range 0.66–2.80 mm). Paired t-test showed a significant difference (t = −3.9, p = 0.001), indicating intraoperative TRE was greater than preoperative TRE (Fig. 4D–E).

Bland–Altman analysis of TRE revealed a mean bias of −0.24 ± 0.37 mm and 95% limits of agreement from −0.96 to 0.48 mm. This indicates that intraoperative TRE was on average 0.24 mm larger than preoperative TRE, which is within clinically acceptable limits and suggests good agreement between the two measurement methods (Fig. 4F).

To investigate differences in intraoperative TRE across spinal regions (cervical, thoracic, lumbar), normality tests were first performed. The cervical group did not meet the normality assumption (p = 0.023); therefore, nonparametric tests were applied. The Kruskal–Wallis H test showed no significant differences among the three groups (p = 0.380). These results indicate that SMR localization accuracy is not related to spinal segment.

Qualitative evaluation

In four patients, pedicle screw trajectories were planned preoperatively using the SMR system. Intraoperatively, the planned paths were visualized as overlays. The surgeon subjectively assessed the trajectories in three dimensions. In three cases, the planned trajectories met the standards and were considered clinically feasible. In one case (25%), a visible deviation of approximately 5 mm occurred. This shift was attributed to slight patient repositioning after model registration due to surgical manipulation, causing misalignment between the model and anatomy.

Learning curve assessment

Using 3D-printed models, three surgeons performed navigation localization. The results are summarized in Table 3.

Table 3.

Comparison of localization results among testers based on 3D-printed models

Variable	Doctor	Mean ± SD (mm)	Range (mm)	95% CI (mm)	χ2	P-value
FRE	Doctor 1	1.48 ± 0.34	0.32–2.51	1.21–1.53	2.14	0.24
	Doctor 2	1.29 ± 0.42	0.29–2.86	1.17–1.46
	Doctor 3	1.33 ± 0.39	0.41–2.73	1.29–1.58
TRE	Doctor 1	1.32 ± 0.26	0.31–2.39	1.32–1.75	1.29	0.39
	Doctor 2	1.28 ± 0.41	0.15–2.43	1.21–1.63
	Doctor 3	1.43 ± 0.31	0.28–2.53	1.29–1.72

Data were tested for normality using the Shapiro–Wilk test and found to be non-normally distributed. Therefore, the Friedman test was used to compare FRE and TRE among three surgeons. Results showed no significant differences in FRE (χ2 = 2.14, p = 0.24) or TRE (χ2 = 1.29, p = 0.39) among the groups. These findings indicate good consistency and robustness of the SMR system across different users. A case–time learning curve was plotted (Fig. 5), showing that after approximately 15 practice cases, the operation time stabilized at around 2.7 min. This further supports the system’s reproducibility and ease of use among different operators.

Fig. 5.

Learning curve for SMR localization. Plot of case number versus operation time for three surgeons. Operation time decreased with increasing practice cases and stabilized at approximately 2.7 min after around 15 cases

Discussion

This study provides preliminary validation of the clinical feasibility and accuracy of a MR-based navigation-SMR. Results show that the system enables rapid vertebral registration and precise localization both preoperatively and intraoperatively, with TRE within clinically acceptable limits. The workflow is simple, the system runs stably, and it shows potential for clinical application.

Advantages of the SMR over conventional navigation

Compared with conventional optical navigation systems, the SMR system does not rely on complex external tracking devices. Instead, it utilizes the built-in SLAM functionality of the HoloLens 2 headset to achieve stable anchoring and interactive visualization of the virtual model in real space. To reduce errors from spinal motion, we adopted a single-vertebra registration approach, effectively minimizing anatomical shifts caused by changes in patient positioning or deformation of the flexible spinal structure. Through a “see-through” navigation interface, the SMR system integrates the virtual model with real anatomy, providing surgeons with an intuitive and immersive spatial localization experience. This not only enhances surgical visibility but also simplifies the workflow, potentially improving efficiency and safety. The wearable, contact-free design offers a lower-cost, more easily deployable navigation alternative for real-world clinical settings.

Comparison with current intraoperative imaging and navigation approaches

Currently, C-arm fluoroscopy is the most widely used intraoperative imaging modality in spinal surgery [18–20]. It offers intuitive localization, relatively low equipment cost, ease of setup, and a short learning curve, making it practical for many routine procedures. However, radiation exposure remains unavoidable, especially in complex or prolonged surgeries requiring repeated imaging, posing health risks to both surgeons and patients [21]. In addition, surgeons must mentally reconstruct three-dimensional anatomy from two-dimensional fluoroscopic images, which adds cognitive load. Previous studies has explored the use of 3D-printed guides for pedicle screw placement, demonstrating feasibility in resource-limited settings [22]. This method preoperatively designs patient-specific guide channels, improving accuracy without the need for advanced equipment. However, 3D-printed guides have clear limitations: once fabricated, they cannot be modified intraoperatively; they lack flexibility for surgical adjustments; and their design, modeling, and printing require considerable time, compounded by the need for sterilization—making them unsuitable for emergency or rapid-response procedures. A previous cadaveric study evaluating MR-based navigation accuracy reported a mean overlay error of 2.1 mm using a manual registration approach based on bony landmarks, demonstrating the feasibility of mixed reality guidance in spine surgery [23]. Building upon this foundation, our study employed a semi-automatic, feature point–based registration method, which achieved TREs below 2 mm in both preoperative and intraoperative settings. Notably, intraoperative TRE was slightly greater than preoperative TRE, likely reflecting the added complexity of the operative environment and subtle surgical field shifts during the procedure.

The SMR eliminates dependence on external tracking hardware by using the HoloLens 2’s built-in depth camera for real-time surgical probe tracking. This significantly reduces system size and improves portability and adaptability in clinical use. Surgeons can independently register multiple target vertebrae based on the actual intraoperative exposure, minimizing registration errors caused by spinal curvature or positional changes. The entire registration process is controlled by voice commands, including landmark selection and registration execution, creating a smoother workflow and reducing the risk of contaminating the sterile field through manual device interaction. The average time to register and localize a single vertebra intraoperatively was approximately 3 min, suggesting that even multiple independent registrations would not markedly prolong navigation time.

Comparison with conventional surgical screw placement techniques

In the context of pedicle screw placement, the current primary methods include intraoperative fluoroscopy and freehand placement. As mentioned earlier, fluoroscopy exposes both the patient and the surgical team to additional radiation, and it does not provide real-time guidance throughout the procedure. Freehand placement, on the other hand, relies heavily on the surgeon’s extensive anatomical knowledge and clinical experience. In this study, qualitative assessments of the pedicle screw placement process revealed that the SMR system offers continuous, real-time visualization of the surgical site without the need for additional imaging. This radiation-free navigation system presents a significant advantage by eliminating radiation exposure, thus minimizing potential health risks for both the patient and the surgical team. Moreover, by incorporating real-time information on insertion depth and angle, the SMR system has the potential to become a powerful assistive tool for pedicle screw placement, further enhancing surgical precision and reducing the likelihood of errors.

Limitations

Despite demonstrating the feasibility of SMR in spinal surgery, this study has several limitations. First, the sample size was small, and there is a lack of large-scale, multicenter clinical data to support the findings. Second, the localization assessment focused only on vertebral surface landmarks and pedicle screw entry points, without addressing the complete pedicle screw insertion process. Real-time guidance of both entry position and insertion angle—critical for clinical application—was not implemented in this study. Additionally, because SMR is point-based, intraoperative identification of vertebral landmarks is required. While generally stable, this method may be limited by inadequate surgical exposure, which can prevent probe contact with certain landmarks. Finally, once registration is completed, the system depends on the absolute stability of the surgical field to maintain correct model overlay. Spinal movement or operating table rotation can introduce localization errors, potentially affecting the accuracy of the procedure.

Future directions

Future plans include expanding the sample size and conducting multicenter trials to validate system stability and generalizability across patient populations and surgeons. For pedicle screw placement, we aim to develop dedicated navigation tools that provide real-time guidance of insertion angles and depths, while employing reference tools and dynamic tracking to reduce registration drift. Hardware improvements will explore lightweight or modular display devices to enhance wearing comfort. Clinically, we plan to establish standardized preoperative planning and intraoperative protocols, and to optimize the interface to reduce the learning curve. Regulatory efforts will focus on obtaining medical device approval, ensuring strict ethical review and informed consent to support safe, lawful clinical deployment.

Conclusion

In summary, this study proposed and preliminarily validated a single-vertebra-based MR spinal navigation system. Utilizing a head-mounted MR device, the system enables rapid preoperative and intraoperative registration and accurate localization, with notable advantages in portability, ease of operation, and visualization. Both laboratory and clinical results demonstrated that the TRE remained within clinically acceptable limits. Without relying on external optical tracking equipment, the SMR system provides surgeons with an intuitive and immersive navigation experience, with the potential to reduce cognitive load, streamline surgical workflows, and enhance safety. Although the study is limited by sample size and functional scope, the findings lay the groundwork for the application of MR technology in spinal navigation and offer a reference for future multicenter validation and functional expansion.

Author contributions

Zhongjie Shi developed the program and drafted the main part of the manuscript. Zirui Su and Lingling Yang contributed to drafting secondary sections of the manuscript and analyzed the research data. Xin Gao, Deyong Xiao analyzed the research data. Yilong Peng and Xiaojun Li translated manuscript. Jianfeng Guo and Shujie Sun evaluated the feasibility of the methods and participated in surgeries. Zhanxiang Wang evaluated the overall feasibility of the study, secured project funding, and provided supervision. All authors reviewed and approved the final manuscript.

Funding

This research was supported by Xiamen Municipal Health and High-Quality Development Science and Technology Program (2024GZL-GG47) and Fujian Provincial Health Commission Young and Middle-Aged Talent Training Project (2022GGB010).

Data availability

All data generated during this study are included in the manuscript. The full software program for the SMR project can be obtained from the corresponding author upon reasonable request.

Declarations

Ethics approval

All patients or their legal guardians provided written informed consent for both the surgical procedures and participation in this study. Consent for the publication of any identifiable images was also obtained from all participants. All procedures involving human participants were performed in accordance with the ethical standards of the institutional research committees, as well as the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. This study was approved by the Ethics Committee of the First Affiliated Hospital of Xiamen University (KYLL202391).

Competing interests

The authors declare no competing interests.