Introduction

Spinal fusion is indicated for severe scoliosis given the risk of continued progression after skeletal maturity [1]. The gold standard to halt curve progression, while providing curve correction includes posterior spinal fusion (PSF) with pedicle screw fixation. However, pedicle screw implantation can be risky given the small diameter of pedicles in severely scoliotic spines or small stature individuals and could result in neurologic, visceral, or vascular injuries to adjacent anatomical structures such as the spinal cord [2]. Various 3D imaging modalities have reduced the use of 2D fluoroscopy in scoliosis surgery, but the navigation process to achieve 3D images such as axial plane images is time-consuming and exposes patients to significant levels of harmful ionizing radiation [1, 3].

Ionizing radiation emitted from radiographic imaging is associated with an increased risk of malignancy and radiation-associated pathologies [3]. A systematic review and meta-analysis found that the incidence rates of cancer, breast cancer and cancer mortality in children and adolescents with scoliosis who had undergone repeated radiographs were significantly higher than in the matched general population [4]. Another study found that patients who had undergone treatment and surgery for adolescent idiopathic scoliosis (AIS) 25 years ago had a five times higher overall cancer rate, than that in the age-matched population [5]. Although intraoperative fluoroscopy-based imaging requires less radiation than CT-based navigation systems, multiple exposures to obtain optimal images can result in increased operative time and consequently, increased radiation exposure [6, 7].

The advancement of machine-vision image-guided surgery (MvIGS) technology in recent years has transformed the paradigm of spinal surgery allowing surgeons real-time imaging with multiple planes of vision and without the use of intraoperative radiation [8]. FLASH™ navigation featuring MvIGS technology utilises structured light imaging by combining a light projector with two stereoscopic video cameras to capture precise and detailed 3D images of the exposed surface anatomy (Fig. 1a and b). This allows the system’s software to integrate the patient’s anatomy with the real-time position of surgical instruments and implants that are utilised during surgery without the need for radiation [8]. Recent studies demonstrated that this technology was able to reduce the total intraoperative radiation time and dose in the adult population compared to another navigation system and 2D fluoroscopy [9, 10]. However, despite its potential advantages, the learning curve associated with this MvIGS has not been evaluated. This present study compares MvIGS to 2D fluoroscopy in a paediatric population undergoing PSF with pedicle screw instrumentation in parallel to the investigation of this technology’s learning curve.

Fig. 1
figure 1

a The FLASH™ navigation featuring MvIGS technology utilises special camera to analyse surface anatomy using only visible light, b Navigation can be performed in under 30 s

Full size image

Materials and methods

Study design & participants

This retrospective review was approved by the institutional review board and performed in accordance with STROBE guidelines. Clinical and radiographic records of paediatric patients (≤ 18 years of age) who had undergone PSF at a tertiary hospital using either MvIGS or 2D fluoroscopy between 2016 and 2021 were reviewed. Procedures before July 2020 were performed using 2D fluoroscopy; procedures during and after July 2020 were performed with MvIGS. Both systems are considered standard of care at this institution. All surgeries were performed by the same surgeon.

Radiologic assessment

All patients underwent pre-operative CT (Siemens Somatom Flash CT scanner) and 2D/3D slot-scanning digital radiography (SSDR) (EOS™ Imaging System) as part of standard care to determine the severity of spinal curvature. Intraoperative 2D fluoroscopy was performed using the Ziehm Vision C-arm (Ziehm, Nuremberg, Germany) platform with an image intensifier. 3D image guidance was performed with the FLASH™ navigation system featuring MvIGS (SeaSpine, California, U.S.A). Radiation doses were expressed as dose area product (DAP) (cGycm2) obtained from radiation dose reports provided by the hospital’s radiology department.

Surgical technique

All patients were positioned prone on a radiolucent operating table, and the spine was exposed using the standard posterior approach. Figure 2 compares the workflow and involvement of intraoperative radiation between MvIGS or 2D fluoroscopy.

Fig. 2
figure 2

Comparison of intraoperative radiation of MvIGS navigation and conventional 2D fluoroscopy

Full size image

In the 2D fluoroscopy group, the entry point was determined using anatomic landmarks. Screw trajectory was then confirmed using fluoroscopy. In the MvIGS group, images from the low-dose CT thoracolumbar spine performed preoperatively were uploaded to the 7D system. The scrub team took approximately 15–30 s to set up basic instruments and up to a minute to set up additional complicated instruments, such as the UTK tracker and Flash Lock. Intraoperatively, the reference frame was clamped to a fixed bony landmark, typically the spinous process, and registration was performed. This took no more than 2 min and can be done for single vertebral levels (segmental registration) or for 2 adjacent vertebral levels (block registration). Should the reference frame be accidentally knocked, a `flash fix’ can be performed without the need for re-registration; this usually takes under 15 s. Once registration was complete, a navigated pedicle probe was used to determine the screw entry point and trajectory. A navigated awl was then used to prepare the entry point. Anteroposterior and lateral fluoroscopy images were then performed to confirm the final positions of the screws.

In all cases in both the 2D fluoroscopy and MvIGS groups, facetectomies were performed only after all screws were inserted. All surgeries from both groups were performed under constant neuromonitoring.

Data collection

Data on demographic, scoliosis aetiology, radiologic and surgical outcomes were collected from patients’ clinical records. Estimated blood loss (EBL) was measured using the gravimetric method. Navigation time for pedicle screw implantation—duration from verifying the accuracy and accepting a registration, to termination of the navigation mode when registering a new vertebra level, or shutting the system down—was captured for the MvIGS group. The time for pedicle screw placement was not available retrospectively for the 2D fluoroscopy group.

Cumulative sum (CUSUM) analysis

The purpose of cumulative sum (CUSUM) analysis is to detect small shifts from the target value [11]. In this study, CUSUM method was used to analyse the learning curve of operative time for 64 cases utilising MvIGS. The CUSUM was calculated as: \(S_{i} = \max \left\{ {0,\left( {S_{i - 1} + \left( {X_{i} - \mu_{0} } \right) - k*\hat{\sigma }_{X} } \right)} \right\}\), where Si = upper CUSUM at time i, Si-1 = upper CUSUM at time (i-1), Xi = observed value at time i, μ0 = the specified target mean, k = specified value of parameter k (0.50), and \(\hat{\sigma }_{X}\) = estimated standard deviation. The CUSUM chart usually runs randomly without any slope. However, a surgeon acquiring a new skill in any surgical procedure would be expected to have an upward slope that will eventually cross a decision interval when the procedure is performed at an undesirable level. When the surgeon masters the skill, the slope will move downward and indicate a decreasing trend. Since the surgeon is competent from the beginning of this study, this point was accepted as the case number where proficiency was obtained.

Statistical analysis

All data was analysed using the SPSS software Version 26 (IBM Inc., Chicago, IL, US). Categorical variables were expressed in counts and percentages (n, %) and compared using the Chi-square test or Fisher’s Exact test; continuous variables were expressed in mean and standard deviation (SD) and compared using the independent-samples T test or One-Way ANOVA. Multiple linear regression was performed to detect associations between hypothesized predictors and EBL. A p value of < 0.05 was considered statistically significant. In Tables 1 through 4, figures in bold represent statistical significance (p < 0.05).

Results

A total of 64 patients who underwent PSF using pedicle screws with 2D fluoroscopy, and 64 patients who underwent PSF using pedicle screws with MvIGS system, were included in this study. Age, gender, BMI, and scoliosis aetiology were comparable between the two groups. Comparing to the 2D fluoroscopy group, there were 23% more patients of Chinese descent, and 11% less patients with Malay descent or other ethnicities in the MvIGS group (p = 0.012) (Table 1).

Table 1 Patient demographic and clinical characteristics between 2D Fluoroscopy and MvIGS groups
Full size table

Improvement over time

In the 64 procedures in which MvIGS was utilized, a learning curve was created by plotting operative time with respect to the number of procedures in order. The maximum CUSUM duration of operation was observed to occur in the 9th case, indicating the operative time for the procedures utilising MvIGS peaked at the 9th case (Fig. 3). As a result, patients in the MvIGS group were divided into two phases based on the temporal relation to the 9th case: Phase 1 group includes cases 1–9 (n = 9); Phase 2 group includes cases 10–64 (n = 55).

Fig. 3
figure 3

CUSUM analysis for the operative time of a single surgeon who had performed 64 cases of PSF with MvIGS. The X axis indicates consecutive cases, and the Y axis indicates the CUSUM score for operative time. Upper control line (UCL) located in the flagged turning point (i.e. 9th case) of curvature indicate the point at which the surgeon transitioned from one phase to another and overcame the operative time-learning curve. UCL Upper control line, Si Upper CUSUM at time i, FlagSi Point falling beyond the upper control limit is flagged

Full size image

Phase 1 signifies the learning period. Comparing to the 2D fluoroscopy group, the change in the largest Cobb angle before and after operation was significantly greater in Phase 1 (33.72 ± 12.99 vs. 43.56 ± 10.89, p = 0.030). In addition, fluoroscopy time was reduced in Phase 1 (37.86 ± 15.95 vs. 27.78 ± 8.32, p = 0.090), while other surgical outcomes remained unchanged (Table 2).

Table 2 Surgical outcomes between 2D Fluoroscopy and MvIGS groups
Full size table

Phase 2 signifies the stage where proficiency is reached. Comparing to the learning period in Phase 1, Phase 2 had lower fluoroscopy time (27.78 ± 8.32 vs. 17.75 ± 5.97, p < 0.001), DAP (82.39 ± 33.45 vs. 48.08 ± 36.95, p = 0.011), navigation time (169.27 ± 72.30 vs. 102.78 ± 38.92, p = 0.025), and EBL (977.78 ± 857.00 vs. 484.37 ± 276.40, p = 0.001) (Table 2 and Fig. 4a). However, navigation time per screw (7.63 ± 2.36 vs. 5.42 ± 3.18, p = 0.054) was not significantly different between the two phases (Fig. 4b).

Fig. 4
figure 4

a Navigation time (minutes) between MvIGS Phase 1 and Phase 2; b Navigation time (minutes) by number of pedicle screws implanted between MvIGS Phase 1 and Phase 2. Note Solid circles depict mean values, Solid lines depict median values, Open circles depict outliers, and Asterisks depict extreme outliers

Full size image

Overall surgical outcomes

As compared to 2D fluoroscopy group, MvIGS reduced fluoroscopy time (37.86 ± 15.95 s vs. 17.75 ± 5.97 s, p < 0.001), DAP (127.27 ± 115.35 cGycm2 vs. 48.08 ± 36.95 cGycm2, p < 0.001), EBL (870.25 ± 616.57 mL vs. 484.37 ± 276.40 mL, p < 0.001), and length of stay (8.31 ± 6.03 days vs. 6.53 ± 2.62 days, p = 0.035) by 53%, 62%, 44% or average blood savings of 385.88 mL, and 21%, respectively. The change in the largest Cobb angles after surgical correction was significantly greater in the MvIGS group by an additional 4% (33.72 ± 12.99° vs. 39.42 ± 11.73°, p = 0.015) (Table 2).

There were no significant changes in the total number of levels that were instrumented (10.00 ± 2.82 vs. 9.69 ± 2.73, p = 0.549), number of pedicle screws inserted (18.30 ± 4.68 vs. 19.80 ± 4.97, p = 0.101), largest pre-operative Cobb angle (59.82 ± 17.20° vs. 63.35 ± 20.51°, p = 0.315), largest post-operative Cobb angle (26.97 ± 14.18° vs. 23.93 ± 14.01°, p = 0.245), and operative time (347.32 ± 84.21 min vs. 342.62 ± 71.32 min, p = 0.749).

Multiple linear regressions

Simple linear regression analyses identified four independent predictors for EBL in MvIGS Phase 2: male gender (β =  − 322.805 ml; 95% CI =  − 31.369, 48.666; p < 0.001), neuromuscular scoliosis (β = 201.119 ml; 95% CI = 20.174, 382.064; p = 0.030), number of levels fused (β = 34.186 ml; 95% CI = 7.904, 60.468; p = 0.012), and operative time (β = 1.667 ml; 95% CI = 0.703, 2.631; p = 0.001). Male gender (β =  − 293.628 mL, 95% CI =  − 438.836, − 148.420; p < 0.001) and neuromuscular scoliosis (β = 179.581 mL, 95% CI = 33,611, 325.551; p = 0.017) were the only variables that remained as significant predictors of EBL in multiple linear regressions (Table 3).

Table 3 Simple and multiple linear regression analyses of patients’ estimated blood loss
Full size table

Surgical outcomes according to scoliosis aetiology

Table 4 shows that compared to patients with AIS, syndromic AIS and juvenile scoliosis, patients with neuromuscular scoliosis had the largest pre-operative Cobb angle (77.75 ± 32.00°, p = 0.041), largest post-operative Cobb angle (34.67 ± 22.05°, p = 0.014), increased DAP (35.98 ± 23.63 cGycm2, p = 0.041), and longer LOS in the hospital (8.08  ± 4.03 days, p = 0.016).

Table 4 Surgical outcomes according to scoliosis aetiology in MvIGS Phase 2
Full size table

Complications

No intra-operative complications were noted. There were no intra-operative neuromonitoring changes in either group, and no neurological complications were reported post-operatively.

Discussion

In this retrospective review, a total of 2,351 pedicle screws were implanted into 64 paediatric patients undergoing PSF utilising FLASH™ featuring MvIGS, and 64 patients utilising 2D fluoroscopy. Not only was the MvIGS system able to significantly reduce intraoperative fluoroscopy time, DAP, EBL, and length of stay by 53%, 62%, 44%, and 21%, respectively, it also gave the surgeon more confidence to correct scoliosis curve by an additional 4%, without increasing the length of the procedure. Our findings also showed that proficiency for operative time using MvIGS in PSF could be achieved at 9 cases.

There was a significant difference in racial groups between 2D fluoroscopy and MvIGS groups. Notably, 92% of the PSF in the 2D fluoroscopy group was performed prior to the COVID-19 pandemic. Although this was not analyzed further, there are three possible reasons for a change in patient’s surgical care-seeking behaviour during the pandemic. First, elective surgeries were postponed by the clinical teams to minimise the risk of viral transmission. Second, children’s parents were apprehensive about viral transmission and chose to postpone their surgery. Third, entry restrictions and border closures may deter international patients from seeking non-emergency medical attention in our centre. Nonetheless, there was a decent racial representation in both groups to allow for generalisability of findings to the whole Singaporean population.

PSF with a segmental pedicle screw concept was first introduced by Suk et al. [12], who reported the idiopathic thoracic curve of 51° in average could be corrected to 16° (69% correction) with a minimum 5-year follow-up. Asher et al. [13] reported on 63% correction using hybrid constructs with hooks, apical sublaminar wires, and pedicle screws, and Cheng et al. [14] reported 68.1% of correction with pedicle screws. Consistently, we observed a 59% of curve correction using 2D fluoroscopy and 63% using MvIGS in both phases. Although curve corrections were increased by 4% by using the MvIGS, this did not significantly increase the overall operative time. We postulate that the additional visualisation allowed the surgeon to identify the entry point and appropriate trajectory faster, eliminating the need for repeated radiation for confirmation.

The time of insertion per screw for the MvIGS group was 5.79 ± 3.15 min per screw, while that for the 2D fluoroscopy group was an average of 10 min in another study [15]. These findings are similar to other studies that found insertion time was lower in navigation groups compared to fluoroscopy-guided groups [15, 16]. However, multiple confounding factors may have influenced the operating time, specifically that we did not capture the time spent inserting pedicle screws in the 2D fluoroscopy group.

Although MvIGS navigation is a novel surgical navigation technology, our study has shown that the average fluoroscopy time and DAP were 17.75 ± 5.97 s and 48.08 ± 36.95 cGycm2, which are generally lower than that of 2D fluoroscopy reported in our study (37.86 ± 15.95 s and 127.27 ± 115.35 cGycm2) or previously published literature (46 s to 69 s and 144.5c Gycm2 to 253.1 cGycm2) [17, 18]. While the fluoroscopy time reported in this study for the MvIGS group was higher (17.75 ± 5.97 s) than other MvIGS studies (4.51 ± 3.72 s) [19, 20], it should be noted that the previous studies were conducted in adult patients undergoing shorter construct procedures with less variable anatomy. Despite the longer fluoroscopy time observed in other studies (80.9 ± 68.1 cGycm2) [9, 20], the DAP reported in this present study remains lower (48.08 ± 36.95 cGycm2). In addition, the use of MvIGS resulted in significant EBL savings of 385.88 mL compared to 2D fluoroscopy. These results are comparable to a recent systematic review which demonstrated that the use of intraoperative computer navigation was effective in lowering intraoperative blood loss, when compared with fluoroscopic guidance [16].

Most AIS patients typically require hospitalisation for 4–9 days for surgery and post-operative medical stabilisation [21]. Length of stay among neuromuscular scoliosis patients is usually longer with previous study showing up to 51 days [22]. Our sub-group analysis reported that patients with neuromuscular scoliosis had the longest length of stay, compared to patients with AIS, syndromic scoliosis or juvenile scoliosis. In addition, curve magnitude was observed to be significantly larger, pre- and post-operatively, in patients with neuromuscular scoliosis. Operative time trended upward, but did not meet a statistically significant relationship with length of stay. Compared to patients with AIS, patients with neuromuscular scoliosis lost more blood intraoperatively. Taken together, our findings are consistent with the current literature that larger and more complex spinal surgeries place more of a physiologic burden on patient and likely contribute to a longer length of stay [24].

This study has some limitations. This was a retrospective analysis, and was vulnerable to selection bias and heterogeneity of the patients studied. Secondly, the accuracy of pedicle screw placement was not validated, as this would require an additional CT scan in a well patient postoperatively. The authors felt this was difficult to justify and it was also beyond the institution’s standard of care. However, pedicle screw placement remained reasonably standardised with a single surgeon, and no signal drop during neuromonitoring or complications pertaining to screw placement were noted in either group. Another limitation is that the collection of radiation doses from imaging device reports might not have been the most precise method to determine patient radiation exposure. In addition, radiation doses were not converted to effective doses owing to a wide margin of error from multiple assumptions and failure of the programme to recapitulate actual patient body sizes [23]. Nonetheless, the radiation dose findings presented in this study are comparable with existing literature [9, 10, 17, 24]. Finally, this study involved patients in adjacent time periods, and while differences in surgical staff may be a confounding factor, the main surgical team, lead surgeon, surgical practice and workflow remained constant during the entire study period, which we believe effectively minimises biasness.

This present study is the first to assess the learning curve for MvGIS in treating paediatric scoliotic deformities in a relatively large cohort of 128 patients. Our findings were in agreement with established studies. In our experience, proficiency with the MvIGS was achieved after 9 cases. Long-term follow-up with postoperative patients can be conducted to evaluate the association between intraoperative DAP and the risk of cancer incidence. Collectively, these would provide further justification for the importance of utilising MvIGS as a safe and accurate alternative to 2D fluoroscopy in paediatric spinal surgery.