Robot‐Assisted Stereotactic Osteotomy Technique for Correcting Thoracolumbar Kyphotic Deformity in Ankylosing Spondylitis: A Technical Note

ABSTRACT

Objective

Precise and reproducible control of wedge resection remains challenging in osteotomy correction for ankylosing spondylitis–related rigid kyphosis, and reports of robotic stereotactic execution beyond pedicle screw placement are limited. This technical note describes the operative workflow of robot‐assisted stereotactic osteotomy (RASO), focusing on quantitative wedge planning and stereotactic execution feasibility.

Methods

The RASO technique and operative workflow are described. Patients with type II ankylosing spondylitis‐related thoracolumbar deformity who underwent single‐level three‐column RASO between May and November 2023 were analyzed. Preoperative planning was performed using Surgimap wedge simulation with ratio‐of‐closure–based prediction, and stereotactic osteotomy trajectories were executed using the Mazor X Stealth Edition system. Perioperative parameters and immediate postoperative radiographic findings were descriptively assessed to evaluate the feasibility and reliability of the proposed operative workflow.

Results

This technique was successfully implemented in 15 patients. The mean operative duration was 304.2 ± 51.4 min, and the mean intraoperative blood loss was 486.7 ± 229.5 mL. Planning–execution concordance demonstrated acceptable agreement, with a maximal ratio‐of‐closure–related deviation of ≤ 5.18° and a mean difference of 0.2° between planned and achieved osteotomy Cobb angles. The mean thoracolumbar kyphosis (TLK) improved from 50.2° to 13.2°, the chin‐brow vertical angle (CBVA) improved from 36.7° to 15.1°, the PT improved from 35.6° to 26.6°, and the sagittal vertical axis (SVA) improved from 207.6 to 93.7 mm. Three intraoperative durotomies occurred during the decompression phase; one patient developed delayed cerebrospinal fluid leakage. No neurological deficits, hardware failure, or mortality were observed. All patients achieved osseous fusion and demonstrated improvements in health‐related quality of life measures.

Conclusions

This technical note demonstrates the technical feasibility and workflow reliability of robot‐assisted stereotactic osteotomy for rigid ankylosing spondylitis–related thoracolumbar kyphotic deformity. Quantitative wedge planning can be reproducibly translated into robotic stereotactic intraoperative execution. Further studies are required to define the broader clinical role and potential applications of this technique.

Trial Registration: Chinese Clinical Trial Registry: 2400090375.

The robot‐assisted stereotactic osteotomy (RASO) technique enables quantitative wedge planning and reproducible stereotactic execution for rigid ankylosing spondylitis–related kyphotic deformity, demonstrating technical feasibility and acceptable perioperative safety.

1. Introduction

Severe thoracolumbar kyphotic deformity secondary to end‐stage ankylosing spondylitis (AS) results in pronounced sagittal malalignment and substantially compromises daily function and overall quality of life [1, 2]. Corrective osteotomy has been demonstrated to effectively restore global alignment and improve functional capacity and health‐related quality of life in this population [3]. However, conventional freehand osteotomy techniques are predominantly experience‐dependent and lack quantitative control of the resection geometry, contributing to the risks of both undercorrection and overcorrection of kyphosis [4, 5, 6].

In recent years, spinal robotic systems—including Mazor X Stealth Edition, ExcelsiusGPS, and TINAVI—have been widely introduced for image‐guided pedicle screw placement, offering enhanced stereotactic precision, reduced reliance on intraoperative fluoroscopy, and mitigation of human‐factor deviation associated with freehand instrumentation [7, 8]. Despite these advantages, current robotic platforms have not demonstrated consistent reductions in operative duration compared with conventional techniques, limiting broader clinical adoption. In addition to screw placement, emerging studies have explored robotic applications in revision reconstruction and complex deformity surgery [9, 10]. In rigid AS‐related kyphosis, stereotactic robotic guidance has the potential to precisely determine the osteotomy plane orientation and enable quantitative wedge resection tailored to predefined alignment targets.

Our team previously introduced the concept of robot‐assisted stereotactic osteotomy (RASO), validated its technical feasibility in a synthetic model and cadaveric spine, and established the corresponding operative workflow parameters [11]. Building upon these preliminary investigations, this technical note presents our initial clinical implementation experience to illustrate a standardized operative protocol and workflow reproducibility. Accordingly, this technical note aims to (i) delineate a standardized and reproducible RASO workflow integrating quantitative wedge simulation with stereotactic robotic drilling execution beyond pedicle screw placement; (ii) summarize early implementation feasibility and perioperative safety, including workflow‐critical intraoperative events and mitigation strategies; and (iii) present planning–execution concordance using ratio‐of‐closure and osteotomy Cobb–based metrics, together with radiological sagittal parameters and CBVA, to support appropriate indications/contraindications and guide future workflow refinement.

2. Methods

2.1. Patient Data

Between May 2023 and November 2023, 15 patients with radiographically confirmed Type II kyphotic deformity [12] were enrolled. The inclusion criteria were as follows: (i) age ≥ 18 years and a diagnosis of AS in accordance with the modified New York criteria; (ii) rigid thoracolumbar kyphosis with lumbar lordosis classified as Type IIA or IIB; (iii) preserved hip joint mobility; and (iv) availability of complete clinical and radiographic records. The exclusion criteria were as follows: (i) unstable Andersson lesions; (ii) thoracic hyperkyphosis > 50°; (iii) severe cardiopulmonary impairment, coagulopathy, psychiatric illness, or other comorbidities interfering with postoperative rehabilitation; (iv) prior spinal surgery; and (v) inability to complete follow‐up or incomplete demographic or imaging data. This study was approved by the Institutional Review Board of our hospital (approval number: S2023‐094‐01) and was registered in the Chinese Clinical Trial Registry (registration number: 2400090375). All procedures were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to surgery, with verbal reconfirmation at the follow‐up under standard privacy disclosure and identity protection protocols.

The baseline demographic variables included age, sex, body mass index (BMI), comorbidities, and deformity subtype. The operative variables included total operative duration, estimated blood loss, staged versus single‐session procedures, and perioperative complications. Patient‐reported outcome measures, including the visual analog scale (VAS) for back pain, the Oswestry Disability Index (ODI), and the Bath Ankylosing Spondylitis Functional Index (BASFI), were collected as part of routine clinical documentation (3‐, 6‐, 12‐, and 24‐month follow‐ups).

2.2. Radiographic Assessment and Robotic System

Full‐length standing spine radiographs and thoracolumbar computed tomography (CT) images were routinely obtained at postoperative and 3‐, 6‐, 12‐, and 24‐month postoperative intervals. Radiographic sagittal parameters included thoracolumbar kyphosis (TLK, T10–L2), lumbar lordosis (LL, L1–S1), pelvic tilt (PT), the T1 pelvic angle (TPA), and the C7 sagittal vertical axis (SVA). The Cobb angle at the osteotomy level (O–Cobb) was defined as the angle between the superior endplate of the cephalad vertebra to the osteotomy level and the inferior endplate of the caudal vertebra. The chin–brow vertical angle (CBVA) [13] was assessed during surgical planning using a standing clinical photograph and full‐body EOS radiograph. All the radiographic measurements were independently performed by two spine surgeons via Surgimap software (version 2.3.1; Nemaris, New York, NY) preoperatively, planned, immediately postoperative, and at the follow‐up, and the mean of the two values was used for analysis. Interobserver reliability, assessed using intraclass correlation coefficients (ICCs) based on a two‐way mixed‐effects model with absolute agreement for single measurements, was excellent, with ICC values of 0.961 for TLK, 0.980 for LL, 0.933 for PT, 0.899 for TPA, 0.987 for SVA, 0.977 for O–Cobb, and 0.967 for CBVA.

Preoperative osteotomy planning in the sagittal plane was executed via the wedge‐based simulation module of Surgimap (Figure 1A). During preoperative simulation, CBVA, pelvic parameters, and global alignment targets were evaluated simultaneously, and the planned osteotomy angle was iteratively adjusted in 5° increments to achieve a balanced compromise between horizontal gaze and global sagittal alignment. The ratio of closure (ROC) described by Diebo et al. [14] was applied to predict target correction, and the postoperative ROC was subsequently calculated to permit comparison with the preoperative simulation.

FIGURE 1.

(A) Preoperative sagittal alignment measurement and osteotomy simulation using Surgimap, demonstrating a planned ratio of closure of 67%:33% at the L2/3 level. (B) Mazor X Stealth Edition (MXSE) planning interface illustrating the RASO trajectory set, matching the simulated sagittal osteotomy angle and forming a fan‐shaped stereotactic distribution within the osteotomy plane. RASO, robot‐assisted stereotactic osteotomy.

Robotic osteotomy planning was performed via the Mazor X Stealth Edition system (MXSE; Medtronic, Dublin, Ireland), which comprises an optical navigation interface, a robotic manipulator, and a C‐arm fluoroscopy unit (GE OEC, Utah, USA). Preoperative lumbar CT scans with a 1 mm slice thickness were uploaded into the MXSE platform to generate a stereotactic osteotomy trajectory identical to that of the simulated Surgimap wedge correction (Figure 1B).

2.3. Statistical Analysis

All statistical analyzes were conducted via SPSS software (version 27.0; SPSS Inc., Chicago, Illinois, USA). The normality of continuous variables was assessed via the Shapiro–Wilk test. The data are expressed as the means ± standard deviations (means ± SDs) for normally distributed variables or as medians with interquartile ranges (medians, IQRs) for nonnormally distributed variables. Between‐time point comparisons were evaluated via linear mixed models, reported as estimated marginal means (EMMs, 95% CIs). Paired Student's t‐tests were used for normally distributed paired data, and Wilcoxon signed‐rank tests were employed for nonnormally distributed data. A two‐sided p value < 0.05 was considered statistically significant.

3. Surgical Technique

3.1. Robot Setup and Surgical Exposure

Prior to anesthesia induction, the MXSE platform was positioned, including attachment of the image adaptor to the C‐arm fluoroscopy unit and fixation of the robotic arm mount. Somatosensory‐evoked potentials, transcranial motor‐evoked potentials, and free‐running electromyography were continuously monitored throughout the procedure. Following induction of general anesthesia, the patient was placed in a prone position on a radiolucent operating table. A standard posterior midline incision was made at the predetermined surgical level. Subperiosteal exposure was carried out laterally to the costotransverse junction at the thoracic region and to the transverse processes at the osteotomy level. The segmental vessels encountered during exposure were controlled with bipolar cauterization and hemostatic packing as needed.

3.2. Registration and Instrumentation

To achieve precise geometric conformity between the osteotomy closure angles and internal fixation, pedicle screw instrumentation was used under robotic navigation at least at the adjacent level. A spinous process clamp was secured to minimize motion transfer discrepancy between the patient and the MXSE system and was subsequently connected to the bone‐mounted bridge. The robotic arm was then referenced to the patient's anatomy, initiating registration. Working volume acquisition was performed via 3D fluoroscopic scanning and point–cloud reconstruction, followed by designation of the region of interest. Two fluoroscopic datasets in anteroposterior and oblique projections were acquired to complete registration. The robotic arm was subsequently activated to execute the preplanned pedicle screw trajectories.

Given the extensive syndesmophyte of AS, segmental pedicle screws were placed at a minimum of five vertebral levels spanning the osteotomy segment. Following robotic instrumentation at the cephalad osteotomy level, preliminary decompression of the L2/3 interspace was performed. The resection included the caudal L2 lamina, bilateral facet joints, and partial cranial L3 lamina. The pedicle bases were skeletonized in the coronal plane to create full visualization of the intended osteotomy base (Figure 2A). After all the exposures and bony preparations were complete, registration was repeated to confirm that the stereotactic accuracy was maintained prior to osteotomy execution (Figure 3).

FIGURE 2.

Diagram of the robot‐assisted stereotactic osteotomy technique. (A) Posterior column decompression with resection of the bilateral pedicles to access the pedicle bases and middle column. (B) Multiple stereotactically aligned drilling trajectories were subsequently unified with a chisel to establish a controlled osteotomy plane in the trans‐intervertebral corridor.

FIGURE 3.

Intraoperative verification of registration accuracy after posterior decompression. (A) Anteroposterior and (B) oblique fluoroscopic acquisitions obtained during repeat registration before stereotactic drilling.

3.3. Robot‐Assisted Stereotactic Osteotomy

Drill trajectory planning adhered to a previously established RASO protocol (Figure 2B) [11]. The key technical principles were as follows: (i) trajectories originated at the pedicle bases, forming a linear osteotomy corridor for subsequent channel advancement; (ii) the drilled channel diameter equaled twice the drill‐guide radius on sequential CT slices to ensure safe margins relative to neural structures; (iii) all bilateral trajectories were aligned sagittally to generate a uniform closure plane; and (iv) trajectory distal tips were oriented toward the anterior vertebral cortex, with adjustment permissible owing to the 30‐mm depth limitation of the standard guide. A 3.0‐mm drill was used for all the stereotactic perforation sequences. The entry points were centered at the pedicle bases, with an angular separation of approximately 7° between adjacent trajectories. A minimum of 12 bilateral vector trajectories were established to fully span the cancellous plane and ensure controlled closure mechanics.

After robotic marking of the bilateral osteotomy boundaries, fluoroscopic confirmation was performed (Figure 4A,B). Once alignment fidelity was verified, the “straight” execution mode was selected within the MXSE interface, enabling continuous sequential drilling for each side with minimized robotic arm repositioning. Upon positional confirmation, drilling was executed through the robotic end‐effector guide, and all trajectories were completed bilaterally. For trans‐intervertebral osteotomy [15], the disk space—frequently ossified in the AS—was debrided via a chisel and rongeurs. Residual endplate structures were prepared, safeguarding neural elements. Ultrasonic osteotomes were employed along the planned wedge perimeter to facilitate block extraction. In cases requiring vertebral column decancellation (VCD), two osteotomy planes were stereotactically created, followed by chisel advancement for segmental release.

FIGURE 4.

Intraoperative view of robot‐assisted stereotactic osteotomy in an illustrative case. (A) Following posterior decompression and pedicle removal, the pedicle bases are fully exposed, and bilateral wires are placed through robotic guidance with (B) radiological confirmation of trajectory alignment. (C) Execution of planned drilling trajectories and stereotactic wedge extraction with preserved dural contours and stable neurophysiological signals. (D) Controlled closure of the osteotomy gap via cantilever correction with satisfactory restoration of the sagittal profile.

Rod contouring was performed strictly according to the Mazor X Align–generated sagittal correction angles. The left cephalad and right caudal rods were provisionally fixed in alternating sequence. A chisel was used to release the remaining cephalad disk attachment and to complete anterior cortex disjunction along the prepared RASO planes. Wedge fragments were longitudinally segmented and removed under direct visualization (Figure 4C). Correction was achieved via controlled manual extension maneuvering, complemented by table adjustment. Final closure of the osteotomy site was performed through cantilever reduction, with compressive stabilization applied across the screws superiorly and inferiorly to the resection site via precontoured rods (Figure 4D). Neurophysiological monitoring was continuous during all phases of correction. Upon confirmation of decompression and the absence of residual osseous or soft tissue impingement, posterolateral bone grafting was performed, a closed‐suction drain was placed, and the wound was closed in anatomical layers.

4. Results

4.1. Surgical Outcomes

Based on the predefined inclusion and exclusion criteria, 15 patients, consisting of 13 males and 2 females, with a mean age of 42.2 ± 10.0 years and a mean BMI of 27.1 ± 4.5 kg/m², were enrolled. The operative and perioperative characteristics are presented in Table 1. The mean operative duration was 304.2 ± 51.4 min, and the mean intraoperative blood loss volume was 486.7 ± 229.5 mL. No intraoperative motor‐evoked potential alterations were observed during deformity correction. Three cases of intraoperative durotomy occurred during the preosteotomy posterior decompression phase. All dural tears were repaired primarily. No immediate postoperative cerebrospinal fluid leakage was observed; however, one patient developed delayed wound healing secondary to cerebrospinal fluid leakage.

TABLE 1.

Demographic and clinical data of the patients.

Case	Gender	Age (years)	BMI (kg/m2)	Comorbidities	ASKD Type	Osteotomy level	Planned ROC	Postop ROC	Operation duration (min)	Intraoperative blood loss (mL)	Postop 72 h drainage (mL)	Complications	Postop‐ hospital stay (day)	Follow‐up period (month)
1	M	38	23.9		IIA	L3	67%:33%	51%: 49%	371	800	450		6	29.9
2	M	35	24.8		IIA	L2	73%:27%	70%: 30%	250	400	560		5	29.8
3	M	36	24.8		IIA	L3	67%:33%	51%: 49%	410	1000	1020	Durotomy, delayed healing	10	29.8
4	F	46	37.2		IIB	L2	73%:27%	53%: 47%	325	500	660		8	29.7
5	M	62	23.0	Osteoporosis	IIA	L2	73%:27%	84%: 16%	300	600	330		7	29.6
6	M	40	28.8		IIB	L2	73%:27%	75%: 25%	245	200	650	Durotomy	9	29.5
7	M	39	22.6		IIA	L2	73%:27%	93%: 7%	232	300	370		6	29.4
8	M	49	32.2	BTHA	IIB	L2	73%:27%	67%: 33%	285	300	520		9	27.6
9	F	64	27.6		IIA	L2	73%:27%	74%: 26%	252	200	520		8	26.2
10	M	38	33.1		IIB	L2	73%:27%	68%: 32%	340	600	560		6	25.2
11	M	41	30.1		IIA	L1	91%:9%	93%: 7%	269	600	450		6	25.0
12	M	38	23.7	Hypertension	IIB	L3	67%:33%	51%: 49%	294	300	550		6	24.0
13	M	36	22.8	Hypertension	IIB	L3	67%:33%	61%: 39%	335	600	400		6	24.4
14	M	26	28.9		IIA	L3	67%:33%	62%: 38%	355	600	630		4	24.2
15	M	45	22.7		IIA	L2	73%:27%	55%: 45%	300	300	700	Durotomy	8	24.0
Mean ±		42.2	27.1						304.2	486.7	558.0		6.9	27.2
SD		10.0	4.5						51.4	229.5	168.2		1.7	2.5

Abbreviations: ASKD, ankylosing spondylitis kyphotic deformity; BTHA, bilateral total hip arthroplasty; BMI, body mass index; ROC, ratio of closure.

In one patient, partial anterior column mobility was identified after posterior decompression and was managed by switching to a dual spinous process clamp at the adjacent level with repeat registration. No robot‐related complications were observed, including registration failure, drill trajectory skiving, intraoperative termination due to system software malfunction, or bone splitting associated with bone‐mounted bridge anchoring. No cases of surgical site infection, neurological deficit, hardware failure, reoperation, or mortality were recorded during the follow‐up period.

Patient‐reported outcomes are summarized in Table 2 and displayed in Figure 5. Linear mixed‐model analysis revealed statistically significant improvements in the VAS score, ODI, and BASFI score at the 3‐, 6‐, 12‐, and 24‐month follow‐up intervals (p < 0.005). Pairwise comparisons between the preoperative and 24‐month evaluations revealed statistically significant improvements across all the scores.

TABLE 2.

Patient‐reported outcome score data at different time points.

Patient report outcome	Preoperative (EMM, 95% CI)	3‐month (EMM, 95% CI)	6‐month (EMM, 95% CI)	12‐month (EMM, 95% CI)	24‐month (EMM, 95% CI)	Pre → 24 m (EMM, 95% CI)	p
VAS	5.60 (4.52–6.68)	2.60 (1.95–3.26)	1.67 (1.02–2.32)	1.40 (0.71–2.09)	1.13 (0.44–1.82)	4.467 (3.660–5.274)	< 0.001*
ODI	41.47 (35.17–47.77)	32.80 (26.16–39.44)	23.07 (17.30–28.84)	19.60 (14.01–25.19)	14.13 (9.52–18.75)	27.333 (21.433–33.234)	< 0.001*
BASFI	5.28 (4.10–6.46)	4.07 (3.19–4.94)	3.51 (2.68–4.33)	3.27 (2.49–4.04)	2.75 (1.90–3.61)	2.527 (1.778–3.276)	< 0.001*

Note: *Significant.

Abbreviations: BASFI, Bath ankylosing spondylitis functional index; CI, confidence interval; EMM, estimated marginal means; ODI, Oswestry disability index; VAS, visual analog scale.

FIGURE 5.

Mean patient‐reported outcomes across five time points: Preoperative, 3‐, 6‐, 12‐, and 24‐month follow‐up. (A) VAS back pain score, (B) ODI score, and (C) BASFI score. BASFI, bath ankylosing spondylitis functional index; ODI, Oswestry Disability Index; VAS, visual analog scale.

4.2. Radiological Results and Correction Evaluation

The radiographic alignment parameters are summarized in Table 3. The average TLK decreased from 50.2° ± 12.3° to 13.2° ± 10.3° postoperatively. Lumbar lordosis (LL) improved from 0.7° ± 24.3° preoperatively to 33.4° ± 19.8° postoperatively. The PT decreased from 35.6° ± 13.4° to 26.6° ± 9.9°. The mean SVA decreased from 207.6 ± 73.4 mm to 93.7 ± 54.4 mm, and the TPA decreased from 49.7° ± 17.7° to 30.3° ± 10.6°.

TABLE 3.

Radiographic measurements of the patients.

Radiographic parameters	Preoperative (mean ± SD)	Plan (mean ± SD)	Postoperative (mean ± SD)	Preoperative to postoperative/P a	Plan to postoperative/P a	24‐month (median, IQR)	Postoperative to 24‐month/P b
TLK (°)	50.2 ± 12.3	23.0 ± 7.2	13.2 ± 10.3	< 0.001*	0.002*	15.2 (8.6–25.8)	0.023*
CBVA (°)	36.7 ± 8.3	12.3 ± 2.6	15.1 ± 2.8	< 0.001*	0.020*	12.6 (12.1–15.7)	0.256
LL (°)	0.7 ± 24.3	39.0 ± 18.0	33.4 ± 19.8	< 0.001*	0.076	37.9 (12.5–47.6)	0.112
O‐Cobb (°)	16.3 ± 11.1	−23.6 ± 11.6	−23.8 ± 13.0	< 0.001*	0.780	−25.0 (−29.5 to −17.9)	0.002*
PT (°)	35.6 ± 13.4	24.3 ± 11.8	26.6 ± 9.9	0.006*	0.374	27.0 (25.0–35.5)	0.012*
TPA (°)	49.7 ± 17.7	25.2 ± 11.9	30.3 ± 10.6	< 0.001*	0.004*	32.6 (20.9–38.9)	0.865
SVA (mm)	207.6 ± 73.4	72.1 ± 59.1	93.7 ± 54.4	< 0.001*	0.040*	86.7 (70.8–152.3)	0.211

Note: *significant.

Abbreviations: CBVA, chin‐brow vertical angle; IQR, interquartile range; LL, lumbar lordosis; O‐Cobb, Cobb angle of osteotomy level; PT, pelvic tilt; SD, standard deviation; SVA, sagittal vertical axis; TPA, T1 pelvic angle; TLK, thoracolumbar kyphosis.

^a Student's t‐tests.

^b Wilcoxon signed‐rank test for nonnormally distributed data.

Assessment of planning accuracy via the ROC method revealed that the immediate postoperative LL, O‐Cobb, and PT values corresponded to the preoperative planning parameters (p > 0.05). In particular, O‐Cobb demonstrated close reproducibility, with a planned mean of −23.6° and an achieved mean of −23.8°, with an absolute deviation ranging from 0.1° to 5.18°. The CBVA decreased from 36.7° ± 8.3° preoperatively to 15.1° ± 2.8° immediately postoperatively (p < 0.001). The planned CBVA was 12.3° ± 2.6°, and the difference between planned and immediate postoperative CBVA was statistically significant (p = 0.020). Maintenance analysis between the immediate postoperative and 24‐month assessments revealed small but statistically significant changes in TLK, O‐Cobb, and PT. CBVA, LL, TPA, and SVA did not significantly differ between the immediate postoperative and 24‐month follow‐up assessments. An illustrative case is provided in Figure 6.

FIGURE 6.

Radiographic series of an illustrative case. (A) Preoperative standing lateral radiograph showing rigid thoracolumbar kyphotic deformity consistent with type IIA classification. Preoperative radiographs showed an SVA of 99.5 mm (blue), and osteotomy was planned at L2/3 with a target O‐Cobb of −21.7° (yellow) and PT of 17.7° (white). (B) Postoperative day‐1 radiograph confirming direct bony contact of the L2 lower endplate and L3 osteotomy surface, with SVA of 20.5 mm, O‐Cobb of −24.1° and PT of 18.8°. (C) Three‐month follow‐up demonstrating intact instrumentation and early cortical consolidation of the osteotomy interface. (D) At the 24‐month follow‐up, complete osseous fusion was observed, and sagittal alignment was generally maintained, with O‐Cobb of −26.2° and PT increased to 26.2°.

5. Discussion

This study describes an initial clinical application of robot‐assisted osteotomy for rigid AS–related kyphotic deformity, with a focus on technical feasibility, operative workflow, and potential pitfalls. Emphasis was placed on stereotactic workflow execution and on practical technical considerations associated with robotic‐assisted osteotomy beyond pedicle screw placement. The operative workflow introduced herein integrates quantitative osteotomy planning with stereotactic robotic execution, enabling predefined wedge geometry with controlled closure ratios. Immediate correction accuracy achieved a maximal deviation of 5.2°, with lumbar lordosis, pelvic tilt, and the osteotomy Cobb angle demonstrating close concordance with the planned values. Perioperative safety was acceptable, with no neurological deficits or robot‐related intraoperative complications, and radiographic realignment and health‐related quality of life measures improved after surgery. Given the high risk of overcorrection in IIA‐type deformities—particularly with postoperative CBVA reduction in patients with preexisting cervical rigidity [12, 13, 16]—precise regulation of the osteotomy angle remains a fundamental objective.

5.1. Indications and Workflow Considerations

The previously reported robot‐related intraoperative complication rates in spine surgery are approximately 3.4%, including a 2.5% rate of robot abandonment [17]. No robot‐related complications were encountered in the present study. Notably, the RASO technique is intended for fixed deformities; one female patient demonstrated partial anterior column mobility at L1/2 following decompression. After motion transfer at the cancellous bone interface was identified, the bone mounting configuration was modified to a dual spinous process clamp at the adjacent level and registration was repeated, restoring stereotactic accuracy without progression to trajectory skiving or procedural interruption. Accordingly, RASO may not be suitable for patients with unstable ALs, incomplete anterior/middle column osteophytes or syndesmophytes unless alternative anchoring constructs beyond spinous process clamping are adopted. When combined with trans‐intervertebral osteotomy, RASO provides an enlarged closure surface defined by the cephalad endplate. In contrast, PSO and VCD require the preparation of two closure planes, prolonging the drilling time due to dual‐plane calibration. For deformities requiring > 40° correction, robot‐assisted VCD remains preferable.

Intraoperative blood loss in this cohort averaged 486.7 ± 229.5 mL, comparable to that previously reported for open single‐level osteotomies (537 ± 121.3 mL) and significantly lower than that reported for dual‐level procedures (1132 ± 417.2 mL) [18]. Hemostasis following pedicle base exposure—through bipolar coagulation and bone wax sealing—appeared to limit cancellous field bleeding despite the creation of stereotactic drill corridors. Blood pressure control was routinely employed. The longest operative duration (410 min) occurred in association with extensive dural repair rather than with robotic execution. Future workflow refinement may include anchoring both the cephalad and caudal vertebrae with multilevel bone‐mounted bridges to consolidate registration into a single cycle.

5.2. Realignment Target and Strategy

Preoperative planning incorporated simultaneous assessment of radiographic parameters and CBVA, with CBVA prioritized within a target range of 10° to 20° due to its established association with horizontal gaze, functional satisfaction, and quality of life in AS [19, 20]. Preservation of postoperative horizontal gaze was therefore prioritized, and limited residual global sagittal malalignment was permitted when necessary. Consistent with previous reports [21], discrepancies between predicted and immediate postoperative CBVA were observed, likely related to residual cervical mobility and patient‐specific postural adaptation. To mitigate the risk of postoperative forced upwards gaze, preoperative EOS imaging and clinical photographs were obtained with patients instructed to assume a maximally flexed cervical posture, thereby introducing a CBVA reserve during planning. Within this framework, postoperative TPA and PT values in the present cohort were largely within or near AS‐specific target ranges [22, 23], which differ from alignment targets commonly applied in degenerative adult spinal deformity [24].

5.3. Perioperative Safety Profile Beyond Robot‐Related Events

No vascular, neurologic, or hardware‐related complications requiring revision occurred, and all patients achieved solid fusion at the follow‐up. Three intraoperative durotomies occurred (20.0%), with one progressing to persistent cerebrospinal fluid leakage and delayed wound healing (6.7%). All dural tears were encountered during posterior decompression and laminectomy. Intraoperatively, dural injuries were observed in regions characterized by severe dural calcification and dense adhesion to the lamina, which are well‐recognized pathological features in patients with long‐standing AS [25]. This interpretation is consistent with previous reports (up to 35.4% in the thoracolumbar and lumbar PSO), indicating that AS represents an independent risk factor for dural injury during spinal osteotomy [25, 26], particularly in the presence of dural ossification and reduced epidural space. Based on these observations, the current results do not support an association between robot‐assisted stereotactic osteotomy and an increased risk of dural injury. Nevertheless, meticulous protection of the dural sac and nerve roots during posterior decompression remains mandatory in AS deformity surgery.

5.4. Technical Context and Future Directions

Historical precedents for precision‐guided osteotomy have focused largely on navigated marking rather than trajectory execution. Metz et al. first reported navigated drilling for osteotomy boundary refinement [27], and Faundez et al. later demonstrated O‐arm–guided osteotomy angle optimization to minimize the risk of pseudarthrosis [28]. Conventional navigation requires repeated gaze diversion from the operative field to the navigation display, whereas the RASO technique permits in‐field visualization, with stereotactic guidance serving as primary vector trajectory control and navigation for redundant safety verification.

Although the procedure was implemented with the MXSE platform, the technique remains scalable to any spinal robot capable of stereotactic drilling. Ou‐Yang et al. demonstrated that screw sagittal angulation and rod contouring significantly influence the magnitude of correction [29]. In this cohort, rod shaping was performed to approximate the robot‐generated plan, although not with quantifiable precision. Computer‐assisted rod bending has been shown to improve alignment fidelity [30]. Currently, Bendini remains the only system enabling intraoperative quantitative rod bending and may integrate into ExcelsiusGPS platforms following acquisition. Future robotic deformity systems must address not only stereotactic osteotomy but also quantifiable implant contouring to potentially preserve planned alignment and long‐term stability [31, 32].

5.5. Strengths

The strengths of this technical note include a standardized, stepwise workflow that extends robotic guidance beyond pedicle screw placement to stereotactic osteotomy drilling, and a quantitative evaluation of planning–execution concordance using ROC‐ and O‐Cobb–based metrics with explicit reporting of CBVA. In addition, intraoperative workflow events and mitigation strategies are transparently documented, which may facilitate reproducibility and safe adoption in appropriately selected rigid deformities.

5.6. Limitations

This study has several limitations. First, the present surgical protocol was restricted to type II AS deformities; patients with Type I or type III sagittal imbalance and severe SVA displacement frequently cannot tolerate prone positioning, and future development of lateral decubitus–based RASO techniques may be needed. Second, although the observed planning–execution deviation was within approximately 5°, the sample size remained limited due to safety considerations and system availability during clinical deployment. Third, all procedures were performed by surgeons who had undergone extensive cadaveric and synthetic model training in robotic osteotomy; therefore, no formal assessment of the learning curve could be conducted. In addition, RASO use is confined to rigid deformities, as current robotic registration and stereotactic tracking frameworks do not support real‐time intraoperative vertebral motion compensation in partially mobile or flexible deformities; residual mobility of the anterior or middle column between the osteotomy vertebra and adjacent segments therefore constitutes a contraindication to the RASO technique. Finally, in single‐level osteotomy for AS–related kyphosis, prioritization of CBVA restoration may inherently constrain global sagittal correction, indicating a potential need for more comprehensive deformity planning strategies in selected cases.

6. Conclusions

Robot‐assisted stereotactic osteotomy enables the translation of quantitative wedge planning into stereotactic intraoperative execution in rigid ankylosing spondylitis–related thoracolumbar kyphotic deformity. This technical note highlights key workflow characteristics, technical considerations, and potential limitations associated with the initial clinical implementation of RASO. Further studies are warranted to refine stereotactic execution strategies and to explore the applicability of this workflow in broader deformity contexts.

Author Contributions

Yi Huang: conceptualization, methodology, formal analysis, software, writing – original draft, writing – review and editing. Yiming Fan: formal analysis, investigation. Han Yu: formal analysis, investigation. Ze Wang: formal analysis, investigation. Zhihao Ma: investigation, visualization, validation. Tianhao Wang: investigation, supervision. Wenhao Hu: investigation, visualization, validation, data curation, resources. Xuesong Zhang: conceptualization, supervision, resources. Guoquan Zheng: conceptualization, supervision. Qi Wang: investigation, visualization, validation. Yan Wang: conceptualization, supervision, project administration, resources, funding acquisition.

Funding

This work was supported by the National Natural Science Foundation of China (No. 82572855).

Ethics Statement

All the authors listed meet the authorship criteria according to the latest guidelines of the International Committee of Medical Journal Editors and are in agreement with the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Huang Y., Fan Y., Yu H., et al., “Robot‐Assisted Stereotactic Osteotomy Technique for Correcting Thoracolumbar Kyphotic Deformity in Ankylosing Spondylitis: A Technical Note,” Orthopaedic Surgery 18, no. 4 (2026): 874–884, 10.1111/os.70282.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.