Novel Machine Vision Image Guidance System Significantly Reduces Procedural Time and Radiation Exposure Compared With 2-dimensional Fluoroscopy-based Guidance in Pediatric Deformity Surgery

Background:

Intraoperative 2-dimensional (2D) fluoroscopy imaging has been commonly adopted for guidance during complex pediatric spinal deformity correction. Despite the benefits, fluoroscopy imaging emits harmful ionizing radiation, which has been well-established to have deleterious effects on the surgeon and operating room staff. This study investigated the difference in intraoperative fluoroscopy time and radiation exposure during pediatric spine surgery between 2D fluoroscopy-based navigation and a novel machine vision navigation system [machine vision image guidance system (MvIGS)].

Methods:

This retrospective chart review was conducted at a pediatric hospital with patients who underwent posterior spinal fusion for spinal deformity correction from 2018 to 2021. Patient allocation to the navigation modality was determined by the date of their surgery and the date of implementation of the MvIGS. Both modalities were the standard of care. Intraoperative radiation exposure was collected from the fluoroscopy system reports.

Results:

A total of 1442 pedicle screws were placed in 77 children: 714 using MvIGS and 728 using 2D fluoroscopy. There were no significant differences in the male-to-female ratio, age range, body mass index, distribution of spinal pathologies, number of levels operated on, types of levels operated on, and the number of pedicle screws implanted. Total intraoperative fluoroscopy time was significantly reduced in cases utilizing MvIGS (18.6 ± 6.3 s) compared with 2D fluoroscopy (58.5 ± 19.0 s) (P < 0.001). This represents a relative reduction of 68%. Intraoperative radiation dose area product and cumulative air kerma were reduced by 66% (0.69 ± 0.62 vs 2.0 ± 2.1 Gycm², P < 0.001) and 66% (3.4 ± 3.2 vs 9.9 ± 10.5 mGy, P < 0.001) respectively. The length of stay displayed a decreasing trend with MVIGS, and the operative time was significantly reduced in MvIGS compared with 2D fluoroscopy for an average of 63.6 minutes (294.5 ± 15.5 vs 358.1 ± 60.6 min, P < 0.001).

Conclusion:

In pediatric spinal deformity correction surgery, MvIGS was able to significantly reduce intraoperative fluoroscopy time, intraoperative radiation exposure, and total surgical time, compared with traditional fluoroscopy methods. MvIGS reduced the operative time by 63.6 minutes and reduced intraoperative radiation exposure by 66%, which may play an important role in reducing the risks to the surgeon and operating room staff associated with radiation in spinal surgery procedures.

Level of Evidence:

Level III; retrospective comparative study.

Posterior segmental spinal instrumentation and fusion are commonly performed for scoliosis deformity correction.1,2 There are inherent risks while implanting pedicle screws due to the adjacent critical neurovascular anatomy. This has led to continuous advancements in intraoperative technology, which have been heavily relied on for guidance in complex cases3Although 2-dimensional (2D) fluoroscopy has historically been the most commonly adopted intraoperative imaging modality, improvements from 2D plain radiographs to 3-dimension (3D) real-time imaging have given surgeons access to multiple planes of vision. Despite the advancements in imaging and navigation, intraoperative radiation exposure remains a concern to surgeons and operating room (OR) staff.

The effects of radiation are especially concerning to the surgeon and staff who are frequently exposed to multiple doses of radiation. This accumulation is also exacerbated as the surgeon must remain in the surgical field, directly adjacent to the image intensifier. Spine surgeons are exposed to 4 to 12 times higher doses of radiation compared with other orthopaedic specialties.4,5 Numerous studies quantify the amount of occupational radiation exposure to health care professionals and provide evidence that occupational exposure is well above the recommended values for annual allowable occupational radiation exposure.5–9 With increased intraoperative radiation exposure, surgeons are potentially subjected to deterministic effects (such as hair loss, skin erythema, skin burns, and cataract formation)10 and stochastic effects (carcinogenesis and teratogenesis).11–13 In a study of female orthopaedic, urology, and plastic surgeons who were exposed to fluoroscopy, it was found that orthopaedic surgeons had twice the expected rate of total cancers and 2.9 times the rate of expected breast cancer.14 Thus, there is a need to develop intraoperative systems that limit radiation to ensure the safety of the surgeon.

A new technology in spinal navigation that utilizes machine vision has been developed to reduce intraoperative radiation to surgeons and OR staff. This system uses advanced optics combining a light projector with 2 stereoscopic video cameras to create a 3D map of the patient’s anatomy and correlates this information with preoperative computerized tomography (CT).15 This system has previously been studied in the adult population undergoing posterior spinal fusion in up to 4 levels and was found to reduce intraoperative radiation compared with both 2D-fluoroscopy-based navigation (2D fluoroscopy) and a different 3D navigation system.16,17 The present study is the first study to compare intraoperative radiation emissions between machine vision navigation to 2D-fluoroscopy in a surgeon population performing spinal deformity correction surgery in a pediatric hospital. The objectives were to compare intraoperative radiation exposure and procedural operative time.

METHODS

This was a single-center retrospective study comparing surgical parameters of pediatric (≤18 y of age) posterior spinal deformity correction surgery with the use of conventional 2D-fluoroscopy-based navigation, OEC Elite (GE Healthcare) to FLASH navigation (SeaSpine, CA). All procedures were performed by a single surgeon with over 20 years of experience in orthopaedic surgery.

Surgeries were performed between 2018 and 2021 and were conducted using either FLASH navigation featuring machine vision or conventional 2D fluoroscopy assistance. Patients were assessed, and treatment was determined by their treating physician as part of their standard of care. Analyzed surgeries were an open posterior fusion of a long construct (6+ vertebras) during a spinal deformity correction surgery. Other procedures (such as transforaminal lumbar interbody fusion) or incomplete data sets were excluded. Navigation modality allocation was determined by the date of the surgery and the date of implementation of FLASH navigation at the participating hospital (Driscoll’s Children’s Hospital, TX). All procedures performed before September 2020 were conducted using 2D fluoroscopy, which was the standard of care at that time. All procedures performed after this date were performed with FLASH navigation, which became the new standard of care.

Institutional Review Board approval was obtained per hospital guidelines to access data from patient’s medical and operative records and discharge summaries, which included: patient demographics, number of vertebrae treated and location (thoracic, lumbar, and pelvis), number of pedicle screws implanted, operative time (defined as the time from the first incision to closure), length of stay, and intraoperative radiation exposure. All intraoperative and postoperative complications were recorded.

Surgical Technique

A standard posterior midline subperiosteal spine exposure was utilized for both imaging modalities. In the 2D fluoroscopy group, the entry point was defined by determining the anatomic landmarks such as the transverse process, and lateral facet border after facetectomy at each level. For thoracic levels, fluoroscopy was used to confirm the entry point. The majority of lumbar pedicle screws were implanted using the freehand technique and without the aid of fluoroscopy. If pedicle probing was difficult due to dense pedicles or abnormal anatomy, fluoroscopy would be used to confirm trajectory. After determining the cancellous part within the pedicle by using a pedicle sounder, sequentially, tapping and screw placement were performed. Anteroposterior and lateral fluoroscopic images were obtained to check whether the screws were positioned properly. All screws were stimulated with a threshold of up to 12 mA for acceptance of the screw.

In the machine vision navigation group, a low-dose preoperative CT scan was obtained and uploaded to the system for each patient. Preoperative CT scans were used for registration of the vertebral anatomy. Facetectomy was done on each level before registration. The surgeon performed segmental registration (registering to a single vertebra) for regions of the spine that were flexible. The reference frame was clamped to the spinous process of the nearest vertebra to place the pedicle screws. By using a navigated pedicle probe, the ideal entry point and trajectory of pedicle screws were defined. Preparing the entry point was done with a navigated awl and probe. The pedicle screws were implanted with the guidance of a virtual trajectory feature (Fig. 1A), allowing the surgeon to follow the trajectory planned using a navigated pedicle probe. The Reslicer (Fig. 1B) was also used during navigation for screw implantation in spinal levels with severe rotation, which allows the surgeon to align the axis marker with the transverse process of the level and create an accurate axial and sagittal view. Anteroposterior and lateral fluoroscopic images were obtained to check whether the screws were positioned properly, and all screws were stimulated similarly to the 2D fluoroscopy group.

FIGURE 1.

A, Live video of the saved augmented reality trajectory, as shown through FLASH navigation. B, The Reslicer feature was used throughout the surgical procedure. It was used to navigate L2 as the axial image resembles an irregular sagittal image (left image) due to severe spinal rotation. The Reslicer is implemented (right image) correcting for the misaligned plane and creating an axial image.

Radiation Exposure

For each case, intraoperative radiation exposure was collected from the radiology department, which was provided on reports from the fluoroscopy system for the duration of the procedure (incision to close). Intraoperative radiation was evaluated by collecting the fluoroscopy system output in dose area product (DAP) in units of gray centimeters squared (Gycm²) and as cumulative air kerma (CAK) in units of milligray (mGy). DAP is a measure of the total energy delivered throughout the entire radiation field and is the product of CAK and the cross-sectional area of the radiation field. CAK is a measure of the energy delivered per unit mass of air to a reference point in the radiation field and is also considered as the scatter radiation as radiation from the source that is deflected off of a surface (such as the patient in an operative setting).18 Scatter radiation exposure is the primary form of exposure to operative staff who stand further away from the surgical table.19

Statistical Methods

Means and SDs are reported for continuous variables; counts and proportions are reported for discrete variables. For continuous variables following approximately normal distributions, the 2 groups were compared using a 2-tailed t test. For discrete variables, a Fisher exact test was used to compare the 2 groups as appropriate. A P value of <0.05 was considered statistically significant. This data set is not available as it includes patient-identifiable information.

RESULTS

A total of 77 children were enrolled in this study: 40 in the 2D fluoroscopy group and 37 in the machine vision navigation group. A total of 1442 pedicle screws were placed: 714 using machine vision navigation and 728 using 2D fluoroscopy. There were no statistically significant differences in the sex, body mass index, or spinal pathologies distributions between groups (P = 0.81, P = 0.81, and P = 0.48, respectively). The mean age was also similar (P = 0.21). No patients received osteotomies or vertebrectomies. All demographic information is presented in Table 1.

TABLE 1.

Patient Demographics and Distribution of Pathologies Underlying Spinal Deformity

	Navigation featuring machine vision (N = 37); n (%)	2D Fluoroscopy-based navigation (N = 40); n (%)	P
Sex
Female	27 (73)	28 (70)	0.81
Age (y)
Mean	13.8	13.2	0.21
Range	9-18	10-17	—
BMI
Underweight	1 (3)	3 (8)	0.81
Healthy	19 (51)	22 (55)	—
Overweight	6 (16)	5 (13)	—
Obese	11 (30)	10 (25)	—
Pathologies
Idiopathic scoliosis	28 (76)	25 (63)	0.48
Neuromuscular scoliosis	7 (19)	12 (33)	—
Congenital scoliosis	2 (5)	2 (5)	—

BMI indicates body mass index; 2D, 2-dimension.

The overall operative time was significantly reduced with machine vision navigation compared with 2D fluoroscopy (294.5 ± 15.5 vs 358.1 ± 60.6 min, respectively, P < 0.001). This is an average reduction of 63.6 minutes or 1.06 hours. The average hospital length of stay trended lower with the machine vision navigation (3.4 ± 0.70 vs 3.8 ± 0.97 d, P = 0.05). Other characteristics of the deformity correction were similar between groups (such as the number of levels operated on, types of levels operated on, and the number of pedicle screws implanted). Characteristics of the procedure are presented in Table 2. No intraoperative complications were reported. Two screws in the 2D fluoroscopy group were repositioned and 1 screw was abandoned and replaced with a pedicle hook based on the neuromonitoring data. No neurological complications were reported postoperatively.

TABLE 2.

Characteristics of Spinal Correction Procedures

	Navigation featuring machine vision (N = 37); n (%)	2D Fluoroscopy-based navigation (N = 40); n (%)	P
Pedicle screws	19.3 (±4.1)	18.2 (±4.4)	0.26
No. levels	13.0 (±2.1)	13.0 (±2.5)	0.64
Levels included in the procedure
Thoracic	37 (100)	40 (100)	1.0
Lumbar	36 (97)	39 (98)	1.0
Pelvis	6 (16)	10 (25)	0.40
Operative time (min)	294.5 (±15.5)	358.4 (±60.6)	<0.001
Length of stay (d)	3.4 (±0.70)	3.8 (±0.97)	0.05

2D indicates 2-dimension.

All intraoperative radiation parameters were significantly reduced with the machine vision navigation group compared with 2D fluoroscopy. The total intraoperative exposure time was significantly reduced with machine vision navigation compared with 2D fluoroscopy (18.6 ± 6.3 vs 58.5±19.0 s, P ≤ 0.001). This represents a relative reduction of 68%. The total intraoperative DAP and CAK were also significantly reduced (0.69 ± 0.62 vs 2.0±2.1 Gycm², P < 0.00001 and 3.4 ± 3.2 vs 9.9±10.5 mGy, P < 0.001, respectively). This represents a relative reduction of 66% in DAP and CAK. Intraoperative radiation parameters are presented in Table 3.

TABLE 3.

Radiation Exposure During Spinal Correction Procedures

	Navigation featuring machine vision (N = 37)	2D Fluoroscopy-based navigation (N = 40)	P
Events	33.6 (±11.6)	104.4 (±32.8)	<0.001
Pedal time (s)	21.6 (±7.8)	71.7 (±25.1)	<0.001
Total exposure time (s)	18.6 (±6.3)	58.5 (±19.0)	<0.001
Total DAP (Gycm2)	0.69 (±0.62)	2.0 (±2.1)	<0.001
CAK (mGy)	3.4 (±3.2)	9.9 (±10.5)	<0.001

CAK indicates cumulative air kerma; 2D, 2-dimension; DAP, dose area product.

DISCUSSION

This is the first study to compare 2D fluoroscopy-based navigation to a novel machine vision navigation system [machine vision image guidance system (MvIGS)] during corrective surgery for a pediatric spinal deformity on the intraoperative radiation of the surgeon and OR staff. These results indicate that the utilization of FLASH navigation featuring machine vision resulted in a significant intraoperative reduction in fluoroscopy time, radiation exposure, and total surgical time.

Historically, 2D fluoroscopy has been a widely utilized method for the guidance of placement of pedicle screws, especially in spines with variable anatomy. Despite its advantages, fluoroscopy-based imaging emits harmful ionizing radiation, to which the risks of iatrogenic radiation exposure have been well-established.20–24 Spine surgeons operating on complex anatomy are particularly susceptible to these effects because of repeated doses throughout a procedure and throughout their career resulting in a significantly higher incidence and mortality from cancer compared with that of the general population.13,14 Intraoperative imaging and navigation have remained a primary source of radiation for spine surgeons and OR staff and warrant a reduced radiation navigation alternative.

Based on this retrospective study involving 77 children and 1441 pedicle screws, machine vision navigation was able to provide 3D intraoperative navigation with reduced intraoperative radiation and operative time. Both modality groups were well matched with no statistically significant difference in age, sex, body mass index, primary pathology, number of operative levels, or number of pedicle screws implanted. Our study found that the intraoperative radiation during spinal deformity correction surgery with machine vision navigation compared with 2D fluoroscopy was exposed to significantly less radiation (DAP and CAK), less total fluoroscopy time, reduced operative time, and a trend of reduced length of stay. This represents intraoperative radiation reductions of 68% and 66%, respectively, and an average time savings of 63.6 minutes. These findings show the substantial advantages and the potential for intraoperative reduced radiation with the implementation of FLASH navigation.

Intraoperative radiation exposure of both modalities used in our study has been previously studied. One study found that 2D fluoroscopy time ranged from 46 to 69 seconds, the DAP ranged from 1.5 versus 2.5 Gycm² and the CAK ranged from 5.1 to 8.8 mGy.25 This range was dependent on the use of a dedicated spine radiology technologist. These values are comparative and validate our findings in the 2D fluoroscopy group (fluoroscopy time; 58.5 s; DAP, 2.0 Gycm²; CAK, 9.9 mGy). A study looking at machine vision navigation in adults undergoing posterior spinal fusion of 1 to 4 levels, found that the fluoroscopy time was 4.51 seconds and the DAP was 0.80 Gycm².17 This represents less fluoroscopy time compared with our study (18.6 s), however, this is likely due to fewer levels fused and less complicated anatomy. The DAP from this previous study is comparable to our findings in the machine vision navigation group (0.69 Gycm²) and is likely because the only potential source of intraoperative radiation when using machine vision navigation is confirmatory fluoroscopic images. Although this is the first study to demonstrate that machine vision navigation reduced radiation exposure in pediatric spinal deformity corrections, our findings are comparable to the existing literature in the adult population.

Furthermore, there was a significant reduction in the total operative time in the machine vision navigation group compared with the 2D fluoroscopy group. The total operative time was reduced from 358.4 to 294.5 minutes or an average reduction of 63.6 minutes. The decreased surgical time in the machine vision navigation group may have been due to increased confidence in the ability to visualize and correct the angle of severely rotated spines leading to a true axial view. It was also due to more confidence in identifying appropriate entry points, and trajectories and understanding the anatomy of dense pedicles. Shorter operative times can deliver advantages including decreased health care costs, decreased potential for surgical and anesthesia complications, and decreased length of stay.26,27 It is suspected that the trend of decreased length of stay in the MvIGS cohort in this study, was related to the significant reduction in total operative time. However, further studies should include patient safety data and other factors that may be correlated to the length of stay. Although limitations exist in quantifying operative time and may include confounding factors such as the effects of rotating staff, machine vision navigation had a significant impact on operative time.

Several limitations exist in the present study. This study was limited by its retrospective nature; thus, it is susceptible to selection bias and group heterogeneity. Inherent to studies spanning multiyear, differences in practice patterns and surgical staff may have resulted in confounding factors, which are not accounted for in the present study, such as different expertise levels, evolving techniques, etc. However, the placement of pedicle screw instrumentation remained reasonably standardized. Pedicle screw placement accuracy was not captured due to ethical reasons. To assess pedicle screw accuracy, the patient would be required to receive additional radiation from a postoperative CT scan, which is not the current standard of care at the institution. Despite not assessing the accuracy, no patients had symptomatic indications of complications relating to pedicle screw breaches, such as neurological deficits. Another limitation of the present study is the impact of coronavirus disease 2019 (COVID-19). COVID-19 has led to numerous cancellations and rescheduled surgeries. Studies suggest that a wait time of 6 months puts patients with idiopathic scoliosis at risk of deformity progression and delaying intervention is associated with prolonged operative times and unfavorable outcomes.28,29 Factors related to COVID-19 were not collected, however, the demographic characteristics suggest that there was no difference between the populations despite machine vision navigation being implemented during COVID-19 in September 2020. There are slightly more neuromuscular patients in the 2D fluoroscopy cohort, though not significant, that may indirectly influence the studied outcomes. In addition, 1 author has an active consultation agreement with SeaSpine at the time of this study, which may have contributed to bias. However, the authors received no financial support for the research, authorship, or publication of this article.

The scope of this study focused on scatter intraoperative radiation that affects the surgeon and OR staff and thus radiation to the patient preoperatively or postoperative has not been directly captured. Scatter radiation is known to be affected by anatomy and the distance of the surgeon and staff to the radiation; thus the actual exposure may vary. Radiation to the patient from the preoperative CT is out of the scope of this study, however, the authors recognize that further investigation is required for patient safety. Preoperative CT has the potential to increase the radiation exposure to the patient depending on the protocol used,30 however, additional radiation exposure to the patient must be weighed with the potential safety advantages provided by the additional 3D data of navigation systems. Low-dose CT scan protocols continue to be experimented with and are crucial in future navigation and radiation studies.

The cost of MvIGS is $500,000 with the addition of disposable passive spheres to navigate instruments (<$200 per case). This is compared with other navigation systems that can cost anywhere from $250,000 to $1,500,000.31 The 2D fluoroscopic system used costs ~$195,000 and does not require preoperative CT imaging. A carbon fiber table or module is required: upgrades to an existing surgical table may cost $30,000 to 60,000 whereas a full carbon fiber table costs ~$100,000 to $200,000. The 2D fluoroscopy system also requires a radiology technician, which has a typical salary range between $56,867 and $66,220.

In this study, the implementation of FLASH navigation yielded considerable benefits for the surgeon performing pediatric spinal deformity corrections. Compared with 2D fluoroscopy-based imaging, total intraoperative fluoroscopy time and intraoperative radiation exposure were all significantly reduced by a relative reduction of 68% and 66%, respectively. Operative time was also significantly reduced by an average of 63.6 minutes. This novel FLASH navigation featuring a machine vision system has reduced intraoperative radiation exposure, thus potentially decreasing the proven harmful effects associated with ionizing radiation toward the surgeon and OR staff.