Image-guided orbital surgery: a preclinical validation study using a high-resolution physical model

Introduction

Orbital pathology is an important part of spectrum of ophthalmic diseases that can be broadly divided into three main categories; neoplastic (50%), inflammatory (25%), trauma and other causes (25%).1 2 These heterogenous pathologies, vary in prevalence according to different age groups.1 2 In infants and up to 2 years of age neoplastic lesions and congenital structural or cystic lesions are more common. In older children from 2 to 16 years of age structural lesions (orbital trauma, dermoid and epidermoid cysts), neoplastic lesions, inflammatory lesions (mostly infectious) and vascular lesions are more common. In individuals aged from 17 to 64 years thyroid associated orbitopathy is by far the most common orbital pathology. Above the age of 65 neoplastic lesions are the most common orbital pathology.2 In this group, there is a high incidence of malignancy and it is imperative for these patients to have focused therapeutic and diagnostic care which often requires complex multi-disciplinary treatment and surgery.

The unique pyramidal shaped orbit is densely packed posterior to the eyeball with sensory and motor nerves, vasculature and extraocular muscles. These delicate structures are enveloped and supported by the orbital septum, which is a fibrous layer arising from the periosteum at the anterior orbital rim as a means of natural protection. There is further cushioning of these structures by unique almost fibrous orbital fat which fills the posterior orbit and is kept separate from anterior orbit by the orbital septum. Therefore, orbital lesions may result in a space occupying mass effect that will cause displacement and/or compression of the orbital contents, neurovascular structures and the eyeball. This manifests as neuro-ophthalmic clinical signs and symptoms such as proptosis, hypoglobus or hyperglobus, impaired ocular motility, diplopia, pupillary abnormalities, reduction or loss of vision, colour vision, and peripheral vision. Thus, orbital pathology and management of these conditions may result in prolonged morbidity3–5 and in some cases the manifested diseases causing mortality.6

The wide spectrum of pathologies faced in this small, dense and complex anatomical landscape makes safe and effective surgical access challenging. Therefore, interventions (ie, biopsy, debulking, resection) may result in complications due to intraoperative damage to the neurovascular structures, the globe and/or the anterior/middle cranial fossa which lie in close proximity. Complications include damage to the globe, optic nerve damage, a permanently dilated pupil and diplopia. One of the most devastating complications following orbital surgery is loss of vision, a serious risk resulting in profound morbidity.6 There are variable reports of blindness following orbital surgery from 0.84% to 7%, with the highest reported with vascular tumours of 4%–7%.3–5 Loss of vision with the latter is often attributable to central retinal artery occlusion secondary to excessive manipulation and bleeding.6

Therefore, advancements in surgical techniques for orbital surgery are required for a more accurate and precise surgical approach to minimise tissue damage and intraoperative complications. There is an imperative to find ways to perform minimally invasive surgery and reduce collateral damage.1 7 Modified surgical techniques have included the use of endoscopic surgery and of navigation image-guided systems (IGS).8 In a similar vein, the fields of neurosurgery and otorhinolaryngology have also looked to advance their surgical techniques by incorporating both of the above techniques.9

With IGS the patients’ actual anatomy is correlated with their preoperative scans via a navigation platform which allows intraoperative real-time tracking of surgical instruments in relation to bony and soft tissue structures. IGS has had its origins in stereotaxy a method developed in neurosurgery that involves the use of external rigid anatomical reference markers for location of internal anatomical surgical landmarks.10 11 There are two main types of IGS navigation systems, optical and electromagnetic, working on two different principles. The optical guided systems uses a light source from infrared or LED cameras that emit beams reflecting off the navigation probes using optical sensors to determine their position within the surgical field. Optical systems determine the position of instruments relative to the patient’s anatomy via these infrared cameras, the static patient reference frame and the specialised tracking instrument. The electromagnetic navigation system on the other hand, tracks instruments and the patient simultaneously and dynamically by generating an electromagnetic field encompassing the anatomical field.10 In electromagnetic image guidance systems (EM-IGS), the operating system is based on a generator that creates an electromagnetic field around the patient’s head which results in a coordinated system within which the reference point as well as the navigating instrument equipped with the built-in electromagnetic sensor are positioned. Spatial movement of the sensor changes the characteristics of the field, and this allows the navigating system to determine the coordinates of the instrument with respect to soft tissue and bone within its field. In EM-IGS platforms, the emitter is fixed to the operating table generating an electromagnetic field around the surgical site, navigation is conducted by a probe or an instrument’s relative position within this field. It is important for orbital surgery that the EM-IGS used are frameless giving unobstructed access to the surgical orbital field and that they do not consist of clamps to immobilise the head or invasive markers attached to the head to assist in navigation which is usually the case with optical systems.

The accuracy of EM-IGS’s surface registration has been questioned in the past as the electromagnetic field is based on narrow field surface registration rendering the depth of orbit distant and that in turn can affect accuracy of navigation, although evidence suggests surface registration is superior to point or fiducial based registration in related anatomical areas.12 13 There is additional interference with ferromagnetic instruments and devices thereby theoretically making the navigation less accurate.

There have been reports on the use of EM-IGS in orbital surgery comparing it to optical IGS, but none have systematically assessed the technology in a preclinical environment to determine if the use of this technology is feasible and consistent in orbital surgery setting. Currently, no bespoke software for orbital IGS exists, and preclinical assessment of the technology’s accuracy within the orbital space has not yet been explored. Therefore, we sought to assess the feasibility and accuracy of EM-IGS using high-fidelity physical orbital anatomy simulators.

Materials and methods

Model creation

In order to assess this orbital specific IGS in a preclinical environment we sought to test its feasibility and accuracy using a physical simulator within a mock operating theatre. However, there is a paucity of validated orbital surgery simulators with the necessary fidelity widely available. Therefore, we developed our own high-fidelity physical models for this study.

Five high-resolution 3D skulls were created with the help of the University College Dublin engineering department specialising in 3D reconstructing and printing. These skulls were printed from radiological scans of five patients with an orbital tumour that previously had a successful orbital biopsy or excision with the help of StealthStation EM-IGS. All patients provided informed consent for the 3D printing of their skulls and were made aware that they would be used for evaluating the accuracy and future training and education of orbital surgeons and trainees. Both male and female patients were asked to be a part of this study to have variations of orbital shape. Each patient had a dedicated CT of the orbits as part of their routine clinical care, acquired with contrast using 64-slice CT Siemens Somatom, at 0.625 mm and reconstructed in three planes using 1 mm slices. The field of view included the tip of the nose to the vertex based on existing sinus navigation protocols. These images were used for both 3D model generation, and for surgical image-guided navigation.

To 3D print the patient models, it is necessary to convert them from a CT format (.DICOM) into mesh format (.STL). First, anonymised patient DICOM scans were filtered using Invesalius 3.1 to highlight the bony structures of the skull and remove all soft tissues. Following this, Autodesk Meshmixer was used to manually remove any unnecessary artefacts from the mesh (such as jewellery and head supports, etc). The model was then sectioned several millimetres above the anterior nasal spine to form a flat base, with all features below this section being removed. It was necessary to reconstruct small sections of the maxilla within the bony orbits as the resolution of the CT scan was unable to detect the thinner sections of bone (see figure 1). This finished model was then exported in STL format for fabrication by 3D printing.

Request permission

Figure 1

(A) Representation of the skull model in Meshmixer with the original scan versus the reconstructed mesh portion within the bony orbit. (B) A preview of the internal printed structure horizontally sectioned along the superior orbital fissure. Slicing was performed using PrusaSlicer. (C) A 3D visualisation of the skull model prior to 3D printing.

The models were prepared for printing using PrusaSlicer 2.4, using a resolution of 0.15 mm layers and a 0.4 mm nozzle. This resolution ensured that the accuracy of the final model was high while keeping print duration within a reasonable timeframe (see representation in figure 1). The print times for each model varied from 47 to 55 hours and were printed with white PLA feedstock on a Prusa Mk3S+ printer. The material extrusion process was used to print in polylactic acid (PLA), which is widely used for its ability to create cost-effective and accurate models. In addition, the white colour and smooth surface of the PLA material also results in an excellent bone-like appearance.

IGS registration assessment

In this study, each 3D skull was placed on a Styrofoam headrest for stability, with the rigid emitter pillow placed under this headrest. A full orbital instrument set was available and relevant instruments were placed in the surgical field to replicate the contemporary theatre environment, and the potential for interference with the electromagnetic field. The thin-sliced CT scans used to generate the 3D models were uploaded onto the IGS ahead of the study.

For IGS to display the instrument’s location in relation to images of the patient, the surgeon creates a map between points on the patient (or model) and points in the uploaded CT/MRI via a process called registration, thereby allowing the system to correlate the preoperative images displayed to the patient (which in this study was the 3D-printed skulls) in real time. Point based surface registration mode was used, requiring the identification of symmetrical reference points involving the orbital rim anteriorly and the tragus posteriorly on both sides. This is done iteratively, until the system has enough points to link the preoperative images to the patient (or model), and in doing so, it produces a system-generated predicted registration accuracy. This is given as an estimation of error in millimetres, with the minimum error driven by the slice thickness of the preoperative images uploaded—in this study, the CT slice thickness was 1 mm, and therefore the minimum achievable registration error was 1.1 mm.

We assessed the ability of ophthalmic surgeons at varying expertise levels to achieve accurate registration in order to explore system feasibility preclinically. Surgeons were selected from the orbital surgery firm within our centre. Each surgeon was asked to attempt registration for each of the five 3D models available, recording their best system generated registration error, and the number of attempts required to achieve this (with a maximum of three attempts allowed). Qualitative feedback on the IGS set-up and registration was also collected. A clinical expert from Medtronic was available throughout the task to explain the registration technique prior to the task, and assist with any technical issues (see figure 2).

Request permission

Figure 2

The process of point based surface registration in green around the two orbits. Preselected symmetrical reference points involved the orbital rim anteriorly and the tragus posteriorly on both sides. Successful registration with a minimum registration error of 1.1 mm was achieved.

IGS navigation assessment

Navigational accuracy was assessed by each participant on each model, using the best registration error achieved on their attempted registration on that model. Navigational accuracy was assessed by placing the navigation probe on defined physical landmarks on the model, and comparing this ground truth with the predicted location of the probe tip on the preoperative imaging, displayed on the IGS display. Landmarks used were on the anterior and apical parts of both orbits: lacrimal fossa, lateral orbital tubercle, supra-orbital notch, lateral edge of superior orbital fissure, and the superior, medial, inferior and lateral points of the optic canal opening. Accuracy was measured by quantifying the difference in the ground truth (ie, probe placed on a defined physical landmark) and predicted probe location (on the IGS display) in millimetres, using inbuilt digital callipers on the IGS (accessed via the ‘Planning’ function). These callipers record a minimum error of 1 mm, and therefore any submillimetre error was not recorded and was deemed sufficiently accurate. Photographic evidence was collected for each measurement for audit purposes (see figure 3).

Request permission

Figure 3

Assessing the navigational accuracy of the system and the instrument at the optic canal. The probe rests on the physical landmark, and the cross hairs on the monitor represents the central tip of the instrument is on point in coronal, sagittal and axial views of the skull. When the probe was placed on the superomedial aspect of the left optic canal, we did not note any discrepancy of the predicted location via the cross hairs on the monitor measured using the inbuilt digital calliper thus demonstrating accurate navigation.

Results

Three independent ophthalmic surgeons participated in the study, one junior trainee, one senior fellow and one consultant. Regarding registration assessment, more senior participants were able to achieve a smaller registration error in a fewer number of attempts (table 1). The junior surgeon achieved a mean best error of 1.14 mm (range 1.1 mm – 1.3), while both the fellow and consultant achieved the minimum possible registration error on each of the five models (mean 1.1, range 1.1–1.1). The junior surgeon required a mean of two attempts (range 1–3) to achieve their best registration error results, while the fellow required 1.6 attempts on average (range 1–2) and the consultant required 1.2 attempts on average (range 1–2). There was a trend towards better registration errors and less attempts needed with consecutive use of the system (table 1), suggesting an underlying learning curve to its use. The limited qualitative feedback from participants suggested a positive experience with the system, highlighting its ‘intuitive’ and ‘user-friendly’ nature.

Regarding navigational accuracy, 16 anatomical points were assessed for each model (eight right orbit, eight left orbit) for each participant. For the majority of points, submillimetre accuracy was achieved and therefore no error was recorded (table 2). The only error was recorded for the junior trainee, using the first model, whereby the best registration error of 1.3 mm was carried forward into the navigation accuracy assessment (table 1). A navigational error of 1.2 mm was noted in assessing the superior, medial, inferior and lateral points of the optic canal opening. This corresponded to the system generated predicted areas of registration inaccuracy at the apex of the orbit, for this particular model and for that particular registration.

More generally, it was noted that the surrounding metallic instruments did not affect the navigation. However, the presence of mobile telephones in close vicinity of the system disrupted and prevented the navigation from working, this was indicated on the monitor by change of colour of the fixation power cross. Removal of such devices from immediate vicinity of the operating table resulted in immediate restoration of navigation with no further need for re-registration.

Discussion

Our study used five patient-specific high-resolution 3D-printed skull models to measure and assess the feasibility and accuracy of EM-IGS navigation in orbital surgery in a preclinical setting. Surgeons of varying expertise level were able to achieve the lowest system-generated prediction registration error (1.1 mm in this study) in the majority of trials. More senior surgeons achieved better registration results, with fewer attempts. Similarly, there was a trend towards requiring less registration attempts with repeated use of the system, suggesting a possible but small learning curve. Furthermore, submillimetre navigational accuracy was achieved for the majority of anatomical points tested, across participants and models. The only navigational error (1.2 mm at the optic canal opening) was recorded for the junior trainee, using the first model, and likely represented carrying forward of a minor registration error. Qualitative surgeon feedback suggested acceptability of the technology, although interference from mobile phones near the operative field was noted. Participants found the system acceptable and usable, with interferences from mobile phones not particularly relevant in a sterile surgical environment. These preclinical results suggest the feasibility of EM-IGS navigation technology in a simulated operation setting.

The potential for this technology is vast, and orbital surgery represents an ideal use case. Safe orbital surgery simultaneously aims to preserve the integrity and function of the globe, extraocular muscles, and the neurovascular structures within the orbit. In recent years, orbital surgery has evolved with the advent of endoscopic approaches for some orbital apex lesions.8 This has improved surgical access but the procedure nevertheless remains challenging and complications persist. Image-guided navigation surgery is widely considered as the next frontier of minimally invasive precise surgery and its use is well established in the fields of neurosurgery and otorhinolaryngology.14 15 There are emerging indications in other fields, for example, orthopaedic limb surgery, dovetailing with the advancement of robotic assisted surgery in these fields.16 The use of IGS in orbital surgery should seem a natural progression as we try to find ways to develop minimally invasive techniques in orbital surgery which fraught with similar challenges that one faces in deep seated neurosurgical and otorhinolaryngology cases.11 13 17 18

This study evaluates the use of EM-IGS in a preclinical fashion through the use of patient-specific 3D-printed skulls with the aim of extending its implementation to orbital surgery. 3D-printed models have several advantages over cadaveric specimens, in that they are cheap, reusable, safe, scalable in size, non-biohazardous, easily stored and disposed of.19 20 3D-printed models have been used for surgical training and preoperative planning; Thawani et al observed that surgeons found 3D-printed models to be more helpful than the virtual models for preoperative planning, diagnosing and treating complex diseases.21 In another study, 3D-printed models of patients undergoing laparoscopic splenectomy were used to preoperatively counsel patients and for subsequent surgical planning.22 They have also been used for reconstruction post-trauma—Park et al described the use of 3D-printed titanium implants for the successful repair of skull defects in 21 cases.23 Similarly, Lin et al looked at 3D-printed skull base tumour model incorporating cranial nerve structures as a neurosurgical training tool, it was observed that these models help facilitate intraoperative navigation and positioning and allowed surgeons to rapidly identify the cranial nerves that are severely displaced by tumours.24

It has been observed in other specialties that IGS proves a useful training tool to teach surgical trainees, highlighting key anatomical structures in realtime.9 25–27 Indeed, we found that orbital surgery trainees in our department reported EM-IGS to be of benefit to their training and understanding of the system as well as surgical planning. This was an added benefit of creating patient specific models, further study in this area would look to validate their use in education, training and technology assessment.

When undertaking literature review, we found that there is a limited preclinical assessment of EM-IGS in orbital surgery. In the available literature, we found that in previous reports of orbital surgery carried out with the help of IGS, the cases selected were particularly challenging and the general observation was that with the assistance of this technology provided high accuracy for targeting the lesion and assisted in anatomic orientation, allowing the surgery to be more controlled and safe.26 28 In patients with orbital apex lesions, this navigational hardware allows for delineation of the optimum trajectory and surgical approach, in addition to precisely localising the tumour that may otherwise be extremely difficult to find.26 A further benefit to EM-IGS is a reported shortening of operative time for complex orbital cases.27–29

In our study, conducted in a preclinical setting, we demonstrated the feasibility and accuracy of orbital EM-IGS navigation system. In our protocol, point-based surface registration points, positioned on both the orbital rim and tragus on both sides, appear to achieve satisfactory navigational accuracy in the tightly packed posterior orbit across all three spatial planes (axial, sagittal and coronal). This is without fixed fiducials, and uses readily available thin slice CT protocols.

We hope that by exploring its accuracy in deep orbit the system will in the future facilitate the precise access to deep-seated lesions in complex orbital cases, contribute to a comprehensive understanding of the adjacent anatomical structures, and heightened awareness of potential intraoperative risks. Indeed this result is in keeping with published literature in neurosurgery and otorhinolaryngology where EM-IGS has been tested for accuracy in preclinical validation studies ex-vivo skull models and in-vivo being found to be accurate and reliable.26 28 30 31

The study possesses several notable strengths, among which the creation of high-fidelity, patient-specific 3D-printed models stands out. Additionally, the study demonstrates a commitment to developing a customised orbital EM-IGS software, tailored to the specific needs of each patient. Another strength of this study lies in its ability to demonstrate the effective integration and simultaneous use of both high-fidelity, patient-specific 3D-printed models and the bespoke orbital EM-IGS software in a simulated operating theatre environment. This showcases the practical feasibility and accuracy of employing these techniques within the orbital context. The limitations of our study included the lack of physical model validation and the small number of surgeon participants.

We hoped that an accurate and feasible orbital EM-IGS navigation system would demonstrate similar levels of accuracy in a real-world intraoperative environment and the data obtained gave us confidence in using and relying on the technology in real-time orbital surgery. This was especially useful in challenging orbital lesions to aid understanding of the surrounding structures, the lesion’s proximity to the surgical site and awareness of the risks involved in intraoperative damage. Our study provides useful a preclinical foundation to planning clinical studies of this technology. This is a necessary first step in the journey of surgical innovation as eloquently put forth ‘the IDEAL pathway of surgical innovation’ which emphasises the need for robust preclinical evaluation (stage 0) study prior to clinical use (stage 1 and beyond).32 Next steps of evaluation would include clinical studies, assessment of learning curves for consultant ophthalmologists and surgical trainees, and cost-effectiveness assessment of the navigation system.

This study demonstrates the accuracy and feasibility of EM-IGS in a preclinical validation study for orbital surgery using patient specific 3D-printed skulls. Additionally, this study demonstrates the successful application of EM-IGS as the navigation system most feasible in orbital surgery with high-fidelity simulation techniques. This preclinical study provides the foundation for clinical studies to explore the safety and effectiveness of this technology and indeed we hope this research will further the application of EM-IGS in orbital surgery.