Interventional Lumbar Spine Procedure Model

Emma Hansen, Quinn Kahle, Aris Magoulas, Ryan Lundeen

Abstract

The accessibility of both materials and easy to use software have led to the widespread adoption of Additive Manufacturing (AM) techniques by hobbyists and industry professionals alike. One application of AM is the quick and low-cost fabrication of limited use, highly customized parts with geometries that are difficult to conventionally fabricate. The medical field has been taking advantage of these characteristics for modeling anatomical structures for demonstration, practice, and surgical uses [1]. The effectiveness of demonstration and practice parts are solely determined by how similar they are to the biologic tissues they are representing. One area that both metal and polymer-based AM techniques struggle in is mimicking the physical properties of biologic tissues [2]. This paper demonstrates that SLA parts printed from unaltered resin can achieve adequate radiopacity for use in surgical simulators.

Introduction

A common AM technique for beginners is Fused Filament Fabrication due to the wide range of shelf-stable materials available, limited post-processing required, and low amount of auxiliary knowledge required to make functional parts. However, FFF parts tend not to achieve adequate HU’s under fluoroscopy.  Several powders have been added to polymer matrices to improve the HU’s of FFF printed parts [1, 3]. However, due to the low supply and prohibitive cost of these doped filaments limiting consistent access and parts produced with these filaments not sufficiently mimicking biological tissues in texture and appearance, doped filaments for the use of increasing a parts HU’s has not been widely adopted [2].

Lumbar spine injections are a fairly common, low-risk procedure used as a pain management technique. The current standard for teaching lumbar spine procedures involves having residents and fellows immediately perform procedures on patients without hand-on practice in a simulated environment [4]. Simulation devices have recently become a viable alternative to train individuals, allowing them to build skills for more complex injections. Simulation devices allow new physicians to become comfortable with the procedure without the risk of learning on patients. These devices typically replicate the lumbar region of the body. Many contain a replica spine, replica cerebrospinal fluid, and replica tissue, as well as circuitry to power various components of the device. However, these devices are fairly expensive, ranging in cost from 1000 to 5000 dollars [5, 6]. Additionally, most current models fail to replicate alternative spinal pathologies, such as scoliosis, which limit their function as an educational tool. It is clear that significant improvements can be made to lumbar injection models to reduce their cost and increase the simulated variability in the educational procedure.

In previous semesters, FFF printing was used to fabricate the five lumbar vertebrae necessary to accurately replicate the lumbar region. The lumbar vertebrae were fabricated by 3D printing a spinal computed tomography (CT) scan that was converted to a 3D design file. Sets of lumbar vertebrae were fabricated with XCT-B Filament, with L1-L5 ranging from 10-40% infill (L1: 10%, L2: 15%, L3: 20%, L4: 30%, L5: 40%). XCT-B filament consists of Acrylonitrile Butadiene Styrene (ABS) filament doped with barium sulfate (BaSO4) [7]. These materials and infill settings were chosen to simulate a variety of radiopacities. A study of the lumbar spine found that L1-L4 have an average radiopacity of 132.6 ± 42.9 Hounsfield Units (HU), providing a benchmark radiopacity value [8]. ABS was used as a negative control for the fabrication procedure, as the material has a known radiopacity of -200 HU at 100% infill [7].

Figure 1: The anatomy at the injection site [3].
Figure 2: Fluoroscopy image of FFF printed lumbar vertebrae. The upper vertebra was determined to look the most realistic to that of cortical bone.

Hounsfield Units

The scale used to measure the relative darkness of tissues in fluoroscopy images is the Hounsfield Unit scale. The Hounsfield Unit scale is defined as follows:

where mu(material) and mu(water) are material specific linear attenuation coefficients (LAC) [9, 10]. An important aspect to note is that the LACs do not consider material thickness or any other manufacturing considerations. While this is not a usually a concern with biological samples, any parts made for surgical simulators will need to account for part thickness and any differences between manufacturing techniques. While the barium sulfate doped ABS had more desirable radiopacity compared to other FFF materials, it was expensive to purchase and lead times for procurement were prohibitive. Therefore, a replacement material was necessary to allow for rapid development at an affordable cost. Liquid resin-based AM techniques have an advantage over solid filaments due to the ease of incorporating additives to improve part performance. For this reason, stereolithography (SLA) was chosen as the new primary manufacturing technology.

Materials

The basic model simplified the anatomical structure of the lumbar spine into 5 different anatomical elements: lumbar vertebrae L1-L5, surrounding tissues (i.e. fat, muscle), ligamentum flavum, spinal canal and skin. The model was broken into these 5 elements as the client identified them as the central anatomical elements that were relevant during lumbar injections. These tissue elements were arranged to mimic the densities of the tissues in the back, without having to recreate the anatomical structure of each tissue layer. Models were fabricated with different material infills to test the optimal model material composition.

The resin used to print the test part and baseline spinal model was the Black V4 from Formlabs. This resin was not doped or modified in any way. It has an ultimate tensile strength of 9380 psi and an tensile modulus of 402 psi, post curing. The heat deflection temperature at 66 psi is 163.6 F post cure and 137.1 F at 264 psi post cure. The baseline spinal printing parameters and results are outlined in the table below.

Table 1: Baseline spinal model component printing parameters and results.

Methods

To determine if SLA printed parts were a feasible alternative to the barium doped FFF vertebrae, a test part was printed and CT scanned. In parallel, a CAD model of three vertebrae (L3, L4, and L5) with incorporated stands and a base were created to be used to correctly place and orient the vertebrae while the simulator was filled with medical gel. The test part was designed to observe the baseline radiopacity of both solid and hollow SLA parts printed with unaltered resin. The part has a triangular shape to view the radiopacity as a function of the thickness of the material.

Figure 3: Dimensions of test part to determine HU of SLA parts.
Figure 4: CAD model of L3, L4, and L5 with incorporated stands and base to be printed using SLA.

The 4 component baseline spinal model, depicted in Figure 4 as a CAD model, was printed using the same material as the test part. As well as the same baseline printing parameters and layer thickness. All 3 vertebrae (L3, L4, L5) being printed on one Form 3 SLA from Formlabs printer given the same printing parameters were calculated to have a 26 hour printing time given no print failure. As a result, L3 and L4 were printed using a Form3 printer from Formlabs and L5 was printed separately on a Form2 SLA printer from Formlabs. There were no printing failures or part warpage as shown in Figure 5.

Figure 5: Printed parts post support removal and processing a) L3 b) L4 c) L5 d) Base

After the lumbar vertebrae and stand were successfully 3D printed, the vertebrae were placed in their proper orientations on the stand. Next, intervertebral felt disks were placed between each vertebra, representing the intervertebral discs. The vertebrae, disks, and stand were placed in the bottom of a pan sprayed with cooking oil to make removal easy. A heat gun was used to melt cubes of medical gel in layers in the pan surrounding the apparatus. Furthermore, polyethylene tubing was threaded through the spinal canal of the model to allow for potential liquid injection practice. The tubing was modified by removing the dorsal portion to allow for injection without being impeded by the tube wall. The opening was covered with cling film such that gel would not be able to drip into the spinal canal after melting. Layers of medical gel cubes were continuously added until the proper coverage was achieved over the vertebrae to create a realistic patient “back”.

Figure 6: Fully fabricated SLA printed lumbar spine model in pan.

The model was tested in an operating room at the Madison Surgery Center using fluoroscopy imaging. Setup for testing was simple. The model was laid flat on and in the middle of the X-ray table. Figure 7 depicts the setup for testing. 

Figure 7: Testing setup at the Madison Surgery Center. The SLA model fabricated this semester is shown in the bottom right corner, along with two other FFF models previously fabricated.

Results

Figures 8 and 9 show the test part in fluoroscopy compared to human hands. While specific HU values were not able to be obtained, the client was satisfied with the darkness of the images.

Figure 8: Comparison of test part compared to human hand in fluoroscopy. View of test part includes air gap and material of varying thickness.
Figure 9: Comparison of test part compared to human hand in fluoroscopy. View of test part shows solid SLA printed material of variable thickness.

The test part demonstrated that SLA printed parts made with unaltered resin can achieve HU values comparable to human bone. A second observation from Figure 4 is that the presence of an air gap, or less than 100% fill volume, does not significantly affect the HU values compared to the 100% infilled view seen in Figure 8. The only criticism made by the client was the lack of definition of the edges of the part. The fluoroscopic image of the final assembly. Like the test part, the client was satisfied with the darkness of the parts but was unsatisfied with the edges of the vertebrae.

Figure 10: Fluoroscopy image of final model assembly.

Conclusions

This project has shown that using a functional test part and baseline human lumbar spine model printed with base SLA process and material that HU units of human lumbar vertebra can be replicated to a high degree, without the need for doping. Along with this the SLA 3D printing process is cost effective and widely available to hospitals and hospital systems. It requires minimal post-processing, which allows for final part production within a day. It can produce a high-resolution model of any patient’s vertebra and can be easily adjusted 3D models allow for more prognoses to be printed with limited CAD experience. Although no doping was required for base SLA material, there is still room to scale structures to better match the Hounsfield units of bone. Doping could still be used for other prognoses and maintain a cost advantage over FFF doped filament.

Future Work

While this project has demonstrated that SLA printed parts using unaltered resin can achieve sufficient HU levels to simulate human bone, there are still several areas to develop. For one, further exploration of the HU equation to account for manufacturing processes and material thickness is vital for achieving consistently accurate simulators. A generalized HU equation that includes factors such as material thickness and manufacturing processes would allow for designers and engineers to adjust part geometry to achieve the desired HU without having to change the base material or printer settings, further easing the burden on the end user.

Another opportunity for further research is in utilizing the various other resins available for use in SLA to model other tissues. For example, the soft rubber material from Formlabs may approximate tendons and ligaments that surround the spine.

A key motivation for this project is to create the ability for a surgeon to make a simulator for any prognosis. This requires the CAD model to be easily modifiable, preferably without the need for dedicated CAD software due to their cost. To this end, developing an STL generator using a common language like Python or C# will allow for an end user to model prognoses like scoliosis specific to each individual patient.

Most importantly, however, further development has to be put into defining the edges of the printed vertebrae. This was the most consistent criticism the client had of our design and procedure. Experimentation has to be done to determine if and how part orientation in the build volume has any effect on the definition of the printed vertebrae under fluoroscopy. Also, post processing of the printed parts such as sanding or sealing need to be explored as well.

References

[1] Trombetta, Ryan et al. “3D Printing of Calcium Phosphate Ceramics for Bone Tissue Engineering and Drug Delivery.” Annals of biomedical engineering vol. 45,1 (2017): 23-44. doi:10.1007/s10439-016-1678-3 

[2] Coles-Black, Jasamine, Damien Bolton, and Jason Chuen. “Accessing 3d Printed Vascular Phantoms for Procedural Simulation.” Frontiers in Surgery 7 (2021). Print. 

[3] M. Savi, M. A. B. Andrade, and M. P. A. Potiens, “Commercial filament testing for use in 3D printed phantoms,” Radiation Physics and Chemistry, vol. 174, p. 108906, Sep. 2020, doi: 10.1016/j.radphyschem.2020.108906. 

[4] “Spinal Injection Simulator,” GTSimulators.com.?https://www.gtsimulators.com/products/spinal-injection-simulator-lf01036u (accessed Dec. 09, 2020). 

[5] “Imaging Course Trainer – Lumbar Spinal Intervention Simulator for Image Guided Injections – 3B?Scientific.” https://www.a3bs.com/image-guided-lumbar-spinal-injection-trainer-p65-1021898-p65-3b-scientific,p_34_30509.html (accessed Dec. 09, 2020). 

[6] V. Manoharan, D. van Gerwen, J. J. van den Dobbelsteen, and J. Dankelman, “Design and?validation of an epidural needle insertion simulator with haptic feedback for training resident anaesthesiologists,” in 2012 IEEE Haptics Symposium (HAPTICS), Mar. 2012, pp. 341–348, doi: 10.1109/HAPTIC.2012.6183812. 

[7] “Filamentos para Impressoras 3D.”? http://www.3don.net.br/Informacoes.aspx?abschnitt=XCT#XCT (accessed May 03, 2022)

[8] “Computed tomography Hounsfield unit–based prediction of pedicle screw loosening after surgery for degenerative lumbar spine disease in: Journal of Neurosurgery: Spine Volume 32 Issue 5 (2020).” https://thejns.org/spine/view/journals/j-neurosurg-spine/32/5/article-p716.xml (accessed Mar. 07, 2021).

[9] Razi T, Niknami M, Alavi Ghazani F., “Relationship between Hounsfield Unit in CT Scan and Gray Scale in CBCT,” J Dent Res Dent Clin Dent Prospects. 2014;8(2):107-110. doi:10.5681/joddd.2014.019

[10] S. Brown, D. L. Bailey, K. Willowson, C. Baldock, “Investigation of the relationship between linear attenuation coefficients and CT Hounsfield units using radionuclides for SPECT,” Applied Radiation and Isotopes, vol. 66,  9 (2008): 1206-1212, https://doi.org/10.1016/j.apradiso.2008.01.002 (accessed Apr. 21, 2022)