Customised hybrid CT-MRI 3D-printed model for grade V spondylolisthesis in an adolescent

Abstract

3D-printed patient-specific models provide added value for initial clinical diagnosis, preoperative surgical and implant planning and patient and trainee education. 3D spine models are usually designed using CT data, due to the ability to rapidly image osseous structures with high spatial resolution. Combining CT and MRI to derive a composite model of bony and neurological anatomy can potentially provide even more useful information for complex cases. We describe such a case involving an adolescent with a grade V spondylolisthesis in which a composite model was manufactured for preoperative and intraoperative evaluation and guidance. We provide a detailed workflow for creating such models and outline their potential benefit in guiding a multidisciplinary team approach.

Background

Three-dimensional printing (3DP) or additive manufacturing has been used in creating patient-specific spine models and instrumentation from clinical imaging data for the past two decades.^{1 2} These models can provide multiple advantages, including visual and tactile sensation, enhanced understanding of spatial relationships and characterisation of anatomical variations and deformities.^3–6 Advances in image processing and coregistration offer the opportunity to create hybrid models with bone anatomy derived from CT and soft tissue/nerve information derived from MRI data. New 3DP technologies can combine materials to produce models with intermediate colours, duraform and textures. Composite models derived from multiple coregistered imaging modalities or sequences are useful in complex surgical planning, for example, in brain tumour resections or vascular neurosurgical cases.^{6 7} 3DP spine models are also helpful for patient education during informed consent of the procedure and explanation of the expected outcome. For surgeons, preoperative models may enable enhanced understanding of patient anatomy and selection of surgical strategy. Intraoperatively, this may lead to reduced surgical time and blood loss with increased accuracy and decreased radiation exposure.^{8 9}

Multimodal 3DP creation is a complex process, requiring several processing steps to create a robust and tangible model. We report the case of an adolescent with an L5/S1 grade V spondylolisthesis where a hybrid CT-MRI composite 3DP model aided in the successful planning and performance of a complicated surgical procedure. We discuss the complete end-to-end workflow, including imaging protocol, quality control and clinical utilisation.

Case presentation

The patient is a 13-year-old woman with no significant previous medical history or family history who presented to our orthopaedic clinic with reportts of mild, intermittent back pain for approximately 2 years, which was exacerbated by a minor ground level fall 4 months prior. After this fall, her pain became constant and was valued from 3 to 7 on a Numeric Rating Scale¹⁰ and responded to tramadol and nonsteroidal anti-inflammatory drugs. Her activity level was significantly compromised as her discomfort increased during participation in recreational activities. Patient also reported of bilateral lower extremity paresthesias and leg pain which extended posteriorly down the legs and noted back pain was worse than the leg pain. No additional sensory reports, motor weakness or bowel/bladder urgency or incontinence was reported.

On examination, a lower lumbar ‘step off’ and mild lower lumbar tenderness to palpation and grossly normal range of motion of the spine was found. No other abnormal findings on comprehensive neurological examination were seen. Patient was consulted on by both the orthopaedic and neurosurgery services and given her clinical picture; a multidisciplinary surgical approach with decompression and stabilisation was recommended. Considering the severity of disease to aid in surgical and implant planning, a 3DP model was ordered.

Investigations

Imaging

The patient was referred to the radiology department for lumbar spine radiography (figure 1A). Lateral radiograph showed grade V anterolisthesis at L5-S1 with bilateral L5 pars interarticularis fractures. L5 vertebra was completely displaced anteriorly as well as inferiorly with respect to the S1 vertebra, with chronic osseous remodelling and sclerosis. L5 and S1 vertebrae were superimposed on the anteroposterior (AP) view, creating the ‘Napoleon hat’ sign (figure 1B).

Figure 1.

Radiographs of the lumbar spine showing grade V spondylolisthesis (arrows). (A) The lateral view demonstrated bilateral L5 pars fractures. The L5 vertebra was completely displaced anteriorly and inferiorly with respect to the S1 vertebra, with chronic osseous remodelling and sclerosis. (B) The anteroposterior view demonstrated superimposed vertebrae, creating the ‘Napoleon hat’ sign on CT.

Images of the lumbar spine were acquired on a Toshiba Aquilion One scanner with zero gantry tilt. Noncontrast scanning was performed using an age-appropriate radiation dose protocol and reconstructed axial slice thickness of 1 mm in axial, coronal and sagittal multiplanar reformats. The CT confirmed chronic bilateral L5 pars interarticularis fractures with corticated fragments, anterior and inferior migration of the L5 vertebral body relative to S1 and anteroposterior widening of the spinal canal. Secondary degenerative changes included loss of disc height at L5-S1 with uncovering of the disc space, degenerative endplate changes and severe bilateral L5 foraminal stenoses—figure 2.

Figure 2.

The sagittal CT scan is shown at midline. (A) Parasagittal views show chronic bilateral L5 pars interarticularis fractures (arrows) with corticated fragments, grade V spondylolisthesis with complete anterior and inferior migration of the L5 vertebral body relative to S1 and anteroposterior widening of the spinal canal (dotted line). (B and C) Secondary degenerative changes included loss of disc height at L5-S1 with uncovering of the disc space, degenerative endplate changes, and severe bilateral L5 foraminal stenoses (asterisks).

MRI was obtained on a 3 Tesla Siemens Prisma scanner, including conventional 2D T1-weighted and T2-weighted sagittal and axial sequences. For 3DP, balanced steady-state gradient echo, T1-weighted and T2-weighted fast turbo spin echo sequences were acquired through the area of interest with isotropic resolution of 1 mm (figure 3). In the setting of patient pain and discomfort, the balanced steady-state gradient echo sequence provided reduced motion while still preserving good definition of the cortical bone and exiting nerves. MRI showed severe compression and deformation of the bilateral exiting L5 nerve roots, with near complete obscuration of perineural fat.

Figure 3.

Isotropic balanced steady-state gradient echo MRI sequence with good definition of cortical bone as well as the exiting nerves was used for 3DP design. (A) Axial and (B) coronal reformats showed severe deformation of the bilateral exiting L5 nerve roots, which appear intermediate in intensity and are compressed by hypointense bony foraminal stenoses (arrows) with obscuration of perineural fat. 3DP, three-dimensional printing.

Computer modelling and 3DP

Process flow for creating the 3D-printed composite spine model is shown (figure 4). All steps were performed under the supervision of a radiologist. Digital Imaging and Communications in Medicine (DICOM) CT images were imported into MIMICS software (Materialise, Belgium). Using tissue segmentation and grayscale intensity thresholding histograms, outlines of the osseous structures including vertebral cortex were created. Region growing then connected all similar-intensity pixels within the stack of images, including any areas of vertebral marrow missed in the first step. Triangulated surface meshes were generated from the data. On MRI, the existing L5 nerve roots were manually segmented. CT and MRI were spatially coregistered using the L5 foramina and additional anatomic landmarks. Composite 3D model of the spine vertebrae and nerve roots were and overlaid on 2D source CT data to perform radiologist quality control before exportation as a stereolithography (STL) format mesh file.

Figure 4.

Workflow for creating the composite 3DP CT-MRI spine and nerve root model. 3DP, three-dimensional printing.

3D printing

Mesh models created in the above manner were printed with a multimaterial printer. We use a Polyjet CONNEX 3 Objet 350 printer (Stratasys, Eden Prairie, Minneapolis, USA), capable of printing simultaneously in multiple colours and material types (hard and flexible materials within the same model). Print resolution and layer thickness settings were X-axis: 600 dpi; Y-axis: 600 dpi; Z-axis: 1600 dpi and 16 microns, respectively. Liquid photopolymer resin jetted out is cured by UV light and solidified; the process is repeated layer-by-layer until the complete model has been printed. Material assignment is critical for proper visualisation. To allow for enhanced visualisation of the stenosed foramina, posterior elements of the spine and the nerve roots themselves were printed in VeroClear and in VeroWhite (Stratasys, Eden Prairie, Minneapolis, USA) photopolymer resin, respectively. After removal of supports in an ultrasonic NaOH (lye) bath, the model was cleaned and coated with polyurethane.

Accuracy of 3DP composite model—reverse validation

To assess the accuracy of the 3DP patient-specific model, the model was similar to the patient (figure 5A). 3D computer mesh model from the images using MIMICS. (figure 5B). The 3D mesh model (target entity) was coregistered to the virtual preprint model (fixed entity) from patient CT 3Matics (Materialise, Belgium) and on the raw patient CT source data (figure 5C) for comparison and to calculate the quantitative spatial variations between the 3D surfaces (figure 6A). Average mean difference between the two models was found to be 0.1 mm±0.1 mm with maximum and minimum values ranging from −0.1 mm to 0.5 mm, respectively (figure 6A, B).

Figure 5.

Image processing of the 3DP spine model is shown, including (A) sagittal CT image, (B) CT scan 3D reconstruction of 3DP model CT and (c) an overlay of secondary virtual model onto raw patient CT data. 3DP, three-dimensional printing.

Figure 6.

Comparison of virtual models generated from 3DP composite model CT and actual patient data, showing spatial variations between the two surfaces. 3DP, three-dimensional printing.

Treatment

Preoperative planning and education using the 3DP model

Preoperatively, the coregistered model was useful for three primary purposes: (1) patient education, (2) planning for the L5 nerve root decompression and (3) planning for placement of the spinal instrumentation. During our preoperative conversations with the patient and her family, the model was instrumental in explaining the deformity and the proposed surgery. Given the degree of changes to the normal anatomy, the surgeons did not believe that the use of a standard spinal model would have conveyed this information as easily or as fully. Preparation for the decompression was facilitated by the spinal model and allowed the neurosurgeon (ES) to plan the approach to the L5 nerve roots, including the amount of bone that would need to be removed and a safe pathway for bone removal. For planning hardware placement, the deformity altered the normal anatomy of the L5 and S1 nerve roots. Having the 3DP model allowed for improved preoperative planning for screw trajectories to achieve compression with the partially threaded cannulated screws. These were placed down the S1 pedicles across the sclerotic changes in the anterior portion of S1 and sclerotic posterior inferior portions of L5 vertebral body to avoid the nerve roots while still allowing for adequate bone purchase.

Surgical plan and execution

Orthopaedic surgery service performed the opening and dissected out all bony anatomy. Using the 3D model as a guide, the cannulated compression screws were placed into the S1 pedicles, traversing into the translated and flexed L5 vertebra. Traditional pedicle screws were then placed bilaterally at the L3 and L4 levels, along with bilateral sacral–alar–iliac screws. The neurosurgery service then performed the bony decompression. For the decompression, a burr drill was used to perform bilateral laminectomy at L5 and this was extended into the inferior aspect of the bilateral L4 lamina. To decompress the L5 nerve roots, a similar technique was used bilaterally: a portion of the L4/5 facets was removed along with a significant portion of the L5 pars interarticularis, which was significantly elongated. After the L5 foramen was identified, the L5 nerve root going into the foramen was also identified and was protected during the case, and no evidence of irritation (as monitored by electromyography) was noted. The L5 nerve root was then fully decompressed by unroofing the posterior aspect of the foramen and then removing a portion of the L5 pedicle to decompress the superior aspect of the nerve. Following these manoeuvres, the L5 nerve roots could be visualised from their exit sites at the thecal sac to beyond the transverse processes. During the decompression, the 3DP model was referenced to better understand the spatial anatomy at the level of the deformity (eg, the elongated L5 pedicle), and the model was found to be highly accurate in regard to the bony entrapment of the exiting nerve root.

Following nerve root decompression, 6.0 mm cobalt chrome rods were placed, the intended fusion sites were decorticated, and allograft and autograft were placed. Postoperatively, the patient had no evidence of neurological dysfunction and was discharged to home on postoperative day 3. Patient was instructed to wear a clamshell lumbosacral orthotic brace when upright.

Outcome and follow-up

Patient had routine follow-up in clinic (last seen 6 months following surgery). Complete resolution of her preoperative pain and paresthesia is observed. Motor and sensory function were completely normal at the most recent follow-up. As seen in post-operative follow-up, anteroposterior and lateral radiographs (figure 7A, B) instrumentation remains unchanged, and patient was released to increased activities without a brace after 6 months.

Figure 7.

Postoperative radiographs (A) lateral (B) anteroposterior—arrow marks the region of interest.

Discussion

3DP spine models have multiple uses including patient and family education, preoperative and intraoperative surgical guidance and medical training. We have described a successful comprehensive workflow for modelling grade V spondylolisthesis using both CT and MRI data for evaluation of bony and nerve root anatomy, respectively. With regard to model accuracy, deviation of the 3DP model from the patient preprint model STL file was maximum and minimum 0.54 mm to −0.14 mm (for a mean of 0.13 mm, an SD of 0.12, a median of 0.12 and a root mean square of 0.18). A previous study using the same polyjet technology for a mandible model showed a maximum and minimum 1.12 to −0.52, mean of 0.17 mm, SD of 0.13 mm and root mean square of 0.21.¹¹

Limitations and challenges of the process are acquisition of high-resolution images’ isotropic MRI data and segmentation of appropriate anatomy. CT and MRI are obtained at different time points, and coregistration of the data sets is a challenging manual process, requiring precise identification of anatomic landmarks on the CT and MRI to reduce errors induced by the process. This can be overcome by placing fiducial markers at strategic anatomic points during the imaging process but requires both CT and MRI to be done at the same visit. Therefore, the process calls for stringent quality assurance processes from imaging to creation of the 3DP model to produce high fidelity models for every patient.

Learning points.

• High-quality three dimensional (3D) image data acquired for complex spinal deformities can generate useful 3D printing (3DP) models with added value for preoperative surgical planning and intraoperative reference.

• Accurate hybrid 3DP models of complex spinal deformities, such as grade V spondylolisthesis, are possible with radiologist-supervised segmentation and coregistration of 3D CT and MRI data.

• 3D-printed models can be useful in counselling/educating patients about their disease process and the surgical steps for correction.