Abstract
Surgery is the mainstay treatment of symptomatic spinal tumors. It aids in restoring functionality, managing pain and tumor growth, and improving overall quality of life. Over the past decade, advancements in medical imaging techniques combined with the use of three-dimensional (3D) printing technology have enabled improvements in the surgical management of spine tumors by significantly increasing the precision, accuracy, and safety of the surgical procedures. For complex spine surgical cases, the use of multimodality imaging is necessary to fully visualize the extent of disease, including both soft-tissue and bone involvement. Integrating the information provided by these examinations in a cohesive manner to facilitate surgical planning can be challenging, particularly when multiple surgical specialties work in concert. The digital 3-dimensional (3D) model or 3D rendering and the 3D printed model created from imaging examinations such as CT and MRI not only facilitate surgical planning but also allow the placement of virtual and physical surgical or osteotomy planes, further enhancing surgical planning and rehearsal. The authors provide practical information about the 3D printing workflow, from image acquisition to postprocessing of a 3D printed model, as well as optimal material selection and incorporation of quality management systems, to help surgeons utilize 3D printing for surgical planning. The authors also highlight the process of surgical rehearsal, how to prescribe digital osteotomy planes, and integration with intraoperative surgical navigation systems through a case-based discussion.
©RSNA, 2024
Test Your Knowledge questions for this article are available in the supplemental material.
Introduction
Spinal tumors may be primary or metastatic in origin. Metastatic spine tumors frequently arising from the lung, breast, prostate, and kidneys are more common than primary spine tumors and occur in 20% of all patients with systemic cancer (1,2). Surgical management is often the mainstay treatment of symptomatic spine metastases and primary spine tumors to achieve improvements in pain, neurologic function, spinal stability, quality of life, and durable local tumor control (3–6). With advancements in medical imaging and three-dimensional (3D) printing techniques, the surgical management of spinal tumors has evolved over the past few decades, enabling a high-precision approach, enhanced safety, and improved outcomes without compromising surgical time.
The ability to create patient-specific 3D printed anatomic models based on imaging data has allowed surgeons to plan and rehearse complex spinal procedures with greater precision and accuracy (7,8). These technologies provide surgeons with a better understanding of the unique anatomy of each patient and, therefore, help to plan the optimal surgical approach. The use of 3D printed models can help identify potential anatomic challenges and guide the selection of the best surgical tools and implants for each patient. The ability to use the patient-specific 3D printed model for surgical rehearsal before the actual surgery provides an opportunity to practice the surgical approach and anticipate potential complications, which can lead to more efficient and effective surgical procedures (7,9).
Another benefit of using 3D printed anatomic models is the potential for improved surgical outcomes. By using 3D models to plan and rehearse the surgical procedure, surgeons can increase the accuracy and precision of their surgical techniques. This can result in reduced blood loss, shorter operative times, lower operating room costs, and fewer postoperative complications (10,11). Outside of the spine, inclusion of tumors within 3D printed models also allows the surgeon to evaluate the adequacy of tumor resection, particularly when the disease is ill defined. This in turn may have an impact on achieving negative margins (12).
In addition to improved surgical planning and outcomes, 3D printed models can also improve patient communication and satisfaction (13). By using a 3D printed model during the discussion regarding surgical treatment, surgeons can help patients better understand the procedure and the risks and benefits associated with it. This can lead to reduced anxiety and decisional conflict and improve patient consent and satisfaction (14). The impacts of 3D printed models on surgical planning, surgical complications, and patient communication and satisfaction are being explored (15–25).
The use of anatomic modeling, 3D printing, and virtual planning in complex spine surgery is still relatively new, and there are some challenges to their adoption. In most cases, the approach to complex spine surgeries requires multimodality imaging including MRI, which best depicts soft-tissue involvement, and CT, which best depicts osseous alterations. Integration of the information provided at CT and MRI may be difficult, particularly in cases where multiple surgical specialties are working in concert. Additionally, there is a need for greater standardization of the design and production process of 3D printed anatomic models to ensure models of high quality that meet safety and regulatory standards.
Despite these challenges, patient-specific 3D printed models and surgical rehearsal are sought-after interventions in complex spine surgery (7,9,26). As 3D printing technology continues to advance and become more widely available, it has the potential to revolutionize complex spine surgery and improve the lives of patients with complex spinal disorders. This article explores the 3D printing workflow, 3D printing materials, quality assurance processes, inclusion of surgical planes, and surgical rehearsal in complex spinal surgeries through a case-based discussion.
3D Printing Workflow
In the following sections, the process of generating 3D printed models from standard-of-care radiologic examinations in patients with thoracic spine tumors is described. Clinicopathologic features and the rationale for the need of a 3D printed model are summarized in Table 1.
Table 1:
Clinicopathologic Features, Surgical Plan, and Outcomes in Thoracic Spine Tumors with Presurgical 3D Printed Anatomic Models
The workflow to generate accurate anatomic models includes patient-specific imaging, segmentation, computer-aided design (CAD) modeling, virtual planning, 3D printing, and postprocessing events (Fig 1). Each of these steps is associated with potential pitfalls that can negatively influence the final 3D printed model. Therefore, a description of how the quality management system is integrated at each stage is included.
Figure 1.
Three-dimensional printing workflow used for creating surgical models from patient data includes Digital Imaging and Communication in Medicine (DICOM) data selection, segmentation, CAD modeling, virtual planning, 3D printing, postprocessing, and quality assurance (QA). QC = quality control.
Image Acquisition and Selection
Imaging of patients with spinal disease includes cross-sectional examinations, such as CT and MRI. The imaging acquisition parameters for CT and MR images used for the creation of 3D printed models are summarized in Table 2. CT and MR images demonstrate complementary information regarding tumor extent. The imaging features often correlate with the underlying histopathologic findings (27–29). CT best demonstrates bone alterations, including osteolytic changes and associated calcifications (Figs 2,3). MRI best demonstrates soft-tissue extent (Figs 2,3). Therefore, often both CT and MRI are used for tumor segmentation (Movie 1).
Table 2:
Thoracic Spine CT and MRI Acquisition Parameters
Figure 2.
Visualization of thoracic spine tumors varies by imaging modality. (A) Sagittal CT image (bone window) in a 63-year-old man (corresponds to case 2 in Table 1) best shows the bone changes (arrow), including cortical erosion of the vertebral body. (B) Sagittal T2-weighted MR image in the same patient best shows the soft-tissue extent (arrow), including the extraosseous mass.
Figure 3.
CT and MRI in a 42-year-old man with a secondary chondrosarcoma arising in an osteochondroma in the thoracic spine (corresponds to case 3 in Table 1). (A) Axial noncontrast CT image of the thoracic spine (bone window) shows a lobular tumor crossing the midline with central chondro-osseous calcification (*). (B) Axial T2-weighted MR image shows the intrathoracic soft-tissue component (arrow), as well as its associated cartilage cap (arrowheads).
Movie 1:
Determination of the optimal imaging examination for tumor segmentation is the initial step in the workflow for 3D printing when using multimodality examinations. MRI and CT demonstrate complementary information regarding the tumor extent. CT best demonstrates the bone changes (arrow), while MRI best demonstrates the soft-tissue extent (rectangle). Therefore, tumor segmentation requires the use of both exams.
Selection of the imaging sequence that best depicts the anatomy of interest is a critical step to achieve an accurate 3D printed model (Figs 4, 5). The MRI sequences are cross-referenced to ensure that the disease and pertinent anatomy are well visualized (Movie 2). Determining the optimal imaging examination, sequence, and series to be used for tumor segmentation is an essential step to achieve an accurate 3D printed anatomic model. Imaging examinations should be reviewed on dedicated reading workstations to identify the anatomy to be segmented and the sequence where it is best visualized.
Figure 4.
Thoracic spine chordoma with various imaging sequences at MRI in a 63-year-old man (corresponds to case 2 in Table 1). Axial T1-weighted (A), T2-weighted (B), and postcontrast (C) lumbar spine MR images show a T2-hyperintense enhancing mass (arrows) originating from the right L1 vertebral body and extending past the midline involving the right L1 pedicle, transverse process, anterior lamina, and facet complex. Epidural extension is present from the T12 to L1–L2 vertebrae with extension into the right T12–L1 foramen and obliteration of the L1–2 foramen.
Figure 5.
Thoracic spine chondrosarcoma with various imaging sequences at MRI in a 45-year-old man (corresponds to case 4 in Table 1). Axial T1-weighted (A), T2-weighted (B), and postcontrast (C) thoracic spine MR images show various features of the left paraspinal mass (*) that abuts the esophagus (★ in C) and extends into the epidural space (dashed arrow in C) and is associated with abnormal bone marrow signal intensity in the T3 vertebral body, T3 pedicle, T4 vertebral body, T4 transverse process (solid arrow) and lamina, and the left fourth rib (arrowhead).
Movie 2:
Once the optimal imaging examinations have been chosen, the optimal imaging sequence for tumor segmentation must be determined. Lumbar spine MRI demonstrates a T2-hyperintense enhancing mass (arrow) originating from the right L1 vertebral body extending past midline involving the right L1 pedicle, transverse process, anterior lamina, and facet complex. Epidural extension is present from T12 to L1-L2 with extension into the right T12-L1 foramen and obliteration of the L1-2 foramen (arrow). Tumor involvement of the right paraspinal muscles is best appreciated with the T2 sequence (arrowhead). Different sequences demonstrate various tumor features. Therefore, segmentation and verification of segmentation accuracy should include these sequences.
The quality of the images directly affects the accuracy of the final 3D model (Fig 6). Imaging resolution can impact the production of the 3D printed model, resulting in a 3D printed model with finer details, smoother surfaces, and a more visually appealing and accurate representation of the anatomy. These differences may be most apparent in the normal anatomy rather than in the tumor due to internal heterogeneous features. Thus, an imaging protocol that prioritizes resolution through smaller section thickness, greater signal-to-noise ratio, contrast-to-noise ratio, spatial fidelity, and optimization of other parameters should be used to create a patient-specific 3D printed model (15).
Figure 6.
Section thickness influences CAD models generated from CT images. CAD 3D model images and corresponding section details in a 15-year-old patient (corresponds to case 1 in Table 1) show that the CAD model generated from DICOM images with a greater section thickness (A) is pixelated and lacks detail when compared with the CAD model generated from DICOM images with a smaller section thickness (B). Inspection of the normal anatomy is critical as detection of variances in pixelation and lack of detail of irregular structures such as the intratumoral chondroid calcifications may be difficult to perceive.
Isotropic voxels where the length, width, and height are identical minimize reconstruction artifacts and are helpful in the evaluation of musculoskeletal abnormalities (30,31). The impact of anisotropic voxels may be minimized when the field of view is small (32). It is of utmost importance to ensure that the 3D printed model depicts every detail possible for optimal surgical planning. Hence, when a 3D printed model is desired, the imaging examination to serve as the data source should be optimized toward 3D printing.
Segmentation
Once the appropriate imaging sequences and series have been reviewed, segmentation using advanced image processing software is performed to outline the anatomies of interest on the Digital Imaging and Communication in Medicine (DICOM) images. Segmentation results in the generation of regions of interest (ROIs) or masks. When both CT and MR images are used for tumor segmentation, a series of steps are first performed before segmentation to ensure the alignment of the anatomy between these examinations, a process known as registration. Registration maximizes the benefits of each modality in the creation of an accurate 3D printed model.
Each software program will manage registration of CT and MR images in their proprietary manner, but the programs commonly require that either CT or MRI is selected as the primary examination onto which the other will be registered. This alignment process is akin to that used when visualizing PET/CT examinations. The software may perform semiautomatic registration (Fig 7) or point-by-point registration.
Figure 7.
Semiautomated registration of overlaid axial MR and CT images at varying opacities in a 15-year-old patient (corresponds with case 1 in Table 1). Images show maximum depiction of MRI findings in color (A), in contrast to the 50:50 blended image of MRI (color) and CT (gray scale) findings (B) and maximum depiction of CT findings in gray scale (C).
The secondary examination is overlaid onto the primary examination and is usually displayed on a color map, the intensity of which can be manipulated to allow optimal visualization of the registered CT and MR images (Fig 7). Each software program manages further refinements of the registration in its own unique manner. Some software programs allow rotation and translation in various axes, while some allow only translation. The refinements to the final registration are verified visually to ensure that the most relevant anatomy is optimally registered. Alignment verification through variability in the intensity of color mapping of the secondary examination, color mapping scale (color), and which imaging examination or sequence serves as the primary target for segmentation is also managed in a software-dependent manner. Several potential problems can arise while performing the registration process. CT and MRI examinations have differences in patient positioning, inspiration, and contrast material timing and possibly different resolutions that can make it difficult to align corresponding structures on images. Optimizing the alignment of the tumor and spine at multiple levels is necessary for accurate registration.
Because the examinations to be registered are obtained with different imaging units and at different time points, motion artifact, including patient movement, degree of inspiration, and orientation during image acquisition, can lead to errors in the registration process. Therefore, alignment of the relevant anatomy such as the tumor and spine are optimized during registration. The alignment of the tumor and adjacent vertebra can be verified at various levels by using tumor landmarks such as calcifications and cystic components. Variance in the alignment of distant anatomy and anatomy influenced by respiration, such as the ribs, heart, and upper abdominal structures, may be expected. Limitations in MRI protocols, absence of voxel isotropy, signal-to-noise ratio, lack of anatomic landmarks, and computational complexities are some of the other potential problems that can arise while registering CT and MR images.
Image resampling or image cropping is one of the effective techniques that can be used to improve registration of CT and MR images. Often CT and MR images can have different fields of view. By cropping the images to include similar anatomic regions, the registration algorithms can focus on aligning the structures that are most relevant. Registration algorithms are computationally intensive, and reducing the size of the images can help improve the speed and efficiency of the registration process. Therefore, image cropping can help reduce the computational burden, improve the accuracy of registration, and reduce the likelihood of errors caused by a mismatch in the anatomic regions included in the examinations.
Segmentation is performed once accurate registration of the examinations is completed and verified. Segmentation of bone findings is often based on the findings at CT, whereas segmentation of soft-tissue findings is often based on those at MRI (Movie 2).
The way these registered imaging examinations are displayed is software dependent. Some software programs allow display of either CT or MR images when performing the segmentation (Movie 3). The resulting mask is tied to the source examination used to generate it. The mask cannot be displayed when selecting the alternate examination or sequence. Software that ties mask display to its source examination or sequence requires alternating repeatedly between the examinations (CT and MRI) or sequences (T2-weighted and postcontrast) to verify segmentation and mask congruence. Alternatively, some software programs allow simultaneous display of the coregistered images and sequences in a new fused series (Movie 4). Segmentation of either bone or soft-tissue findings is possible in this single fused series. This method allows nearly seamless segmentation verification and validation of mask congruence.
Movie 3:
The workflow for segmentation using multimodality examinations includes image registration, segmentation, and verification of the segmentation. CT-MRI registration workflow process depict the imaging exams following registration. Disease extent depicted with MRI is color coded in blue, while disease extent depicted with CT is color coded in pink. Verification requires overlapping disease extent from MRI and CT to ensure disease is adequately represented.
Movie 4:
Fusion is an alternate display format following image registration that allows visualization of both examinations simultaneously. CT and MRI may be registered and displayed as a fused image. Once the examinations are registered, alignment may be optimized through rotation or translation. CT and MRI displayed as a fused image allows verification of alignment and segmentation accuracy (asterisk) both for MRI and CT while in the same window. Posterior translation to optimize alignment of the thoracic spine (arrows) can be easily verified in the same window by changing the display intensity. This allows visualization of the soft-tissue structures, tumor (asterisk), and bones with an easy transition between the CT and MRI examinations.
Thresholding, region growing, edge detection, and semiautomatic algorithm–based tools are generally used to perform the initial segmentation. Manual contouring is often used to refine the initial segmentation or when the segmentation target is complex, irregular, or difficult to distinguish from the surrounding tissues. Once the segmentation is complete, masks of the segmented anatomies are visually inspected. The 3D renderings of all the masks are overlaid on the images for verification of segmentation regardless of the imaging modality source (Fig 8).
Figure 8.
Image display following registration using CT as the primary imaging examination in a 45-year-old patient (corresponds to case 4 in Table 1). (A–C) Segmentation and resulting ROIs of various anatomy including the arteries (red), veins (blue), esophagus (teal), tumor (purple), and paraspinal vessels (pink) are reviewed on axial (A), coronal (B), and sagittal (C) CT images. (D) Once segmentation is complete, the corresponding parts are generated and reviewed on a 3D rendering. (E) Axial fused image of the registered CT and MR images permits visualization of segmented anatomy regardless of the imaging modality source, allowing seamless verification of segmentation.
At our institution, segmentations and 3D renderings are reviewed by a radiologist to ensure the accurate inclusion of the disease due to the complex nature of the cases referred for the creation of 3D printed models. Inspection is performed in all three orthogonal planes, as masks may appear accurate along one plane but nonphysical artifacts often related to imaging technique may be present in other orthogonal planes (Table 3).
Table 3:
Roles within the 3D Printing Multidisciplinary Team
3D Printing Material Selection
Choosing the right material to represent the different anatomies in the 3D model is another essential step to achieve both a functional and visually appealing 3D printed model. Color, texture, and transparency are the three factors to be considered when choosing a material for 3D printed models. Achieving a realistic texture is often important to ensure that the model accurately represents and mimics the human anatomy, as organs have different tactile properties. Materials with matching properties to those of the anatomy should be chosen for a better surgical rehearsal experience. For example, materials used to 3D print the skeletal system must be strong and durable, whereas materials used to 3D print vasculature or muscles should be soft and flexible.
The type of material selected can enhance the visibility of disease in difficult-to-visualize areas or can resemble tissue properties during simulation of surgical maneuvers. Using transparent or clear materials when 3D printing bones such as the ribs and vertebrae allows visualization of epidural disease and intraosseous and bone marrow involvement when disease is depicted in color (Figs 9, 10). Additionally, 3D printing soft-tissue anatomies including the aorta, esophagus, and veins using flexible materials in different colors allows clear visual differentiation among adjacent anatomies while simultaneously mimicking the properties of the native tissues (Fig 10). The utilization of tissue-mimicking materials allows simulation of surgical maneuvers. Therefore, at our institution, we choose transparent rigid resins for skeletal components and colored rigid resin for tumors and blend colored and flexible resins for soft tissues (aorta, esophagus, etc) when using PolyJet (Stratasys) 3D printers.
Figure 9.
Use of clear material for a 3D printed model in a 63-year-old man with thoracic spine chordoma (corresponds to case 2 in Table 1). Lateral (A) and craniocaudal (B) photographs of the 3D printed model show the spine (clear) and tumor (yellow). Use of the clear material allows visualization of the intraosseous and epidural extent of the disease (arrows).
Figure 10.
Various material types to mimic anatomic properties and to allow rehearsal of surgical maneuvers before surgery (corresponds to case 4 in Table 1). Frontal (A), lateral (B), and posterior oblique (C) images of a 3D printed model show the bones (clear), tumor (*), and paraspinal vessels (arrowheads in B) as a single multicolor part printed with rigid materials. The use of clear material for the bones allows visualization of the intraosseous extent of the disease (arrows in C). Flexible material was used to mimic the anatomic properties of the aorta (pink in A and B), esophagus (★ in A and B), and vein (red). Mixed rigid and flexible materials were used to create the trachea (yellow) and mimic its anatomic properties.
Part design is another important element in the creation of 3D printed anatomic models. Materials, including color and texture, are important in the representation of anatomy, but manipulation of these parts can be enhanced by support structures and connectors. This is of importance when multiple surgical specialties are involved in a single case. The use of magnets and removable parts allows manipulation of individual anatomies that each subspecialty will manage and visualization of the anatomy of greatest interest (Movie 5).
Movie 5:
The 3D printed model includes the bones (clear), tumor (asterisk), and paraspinal vessels (arrowheads) as a single multicolor part printed with rigid materials. Use of the clear material for the bones allows visualization of the intraosseous extent of the disease (arrows). Flexible material was used to mimic the anatomic properties of the aorta (pink), esophagus (star), and vein (red). Mixed rigid and flexible materials were used to create the trachea and mimic its anatomic properties (yellow). As some of the structures were to be mobilized during surgery, these were designed to be removable and were connected through magnets (arrows). Material selection mimicking anatomic properties allows for rehearsal of surgical maneuvers prior to surgery.
Quality Assurance
Quality assurance (QA), quality control (QC), and quality improvement are critical in 3D printing at point-of-care facilities, especially when it comes to the production of anatomic models that will be used in surgical planning and patient and trainee education. Developing, utilizing, and revising standardized protocols for 3D printing, postprocessing, and QC enables production of accurate, precise, safe, and highly reliable anatomic models. QA and QC must be performed at every step of the workflow (33,34). Verification during multiple steps of the 3D printing workflow is necessary for accuracy of the 3D printed model and prescribed surgical planes.
QA begins with evaluation of the imaging examinations to determine image quality. The resolution of the images and the presence of imaging artifacts directly impacts the quality and accuracy of the 3D printed model (34). Standardized segmentation protocols using software with U.S. Food and Drug Administration 510(k) clearance must be established for each anatomic region. Segmentation protocols usually designate segmentation parameters, such as threshold settings, region growth, smoothing filters, and active contour segmentation. The protocol should be consistent and must be followed consistently (Fig 11). Once the segmentation is complete, the masks must be validated and verified by an imaging expert or radiologist. It is important to document the segmentation process, including the segmentation parameters and QC measures used, and the validation results (Fig 11).
Figure 11.
QC before 3D printing ensures that the segmentation and corresponding ROIs accurately reflect DICOM source data (corresponds to case 3 in Table 1). (A–D) Documentation log (A) shows specific segmentation parameters used to segment or create each ROI. Axial (B, C) and sagittal (D) CT images show how segmentation verification is performed by obtaining measurements of the anatomy on the DICOM images at three points: anteroposterior dimension of the T2 vertebral body (B), transverse dimension of the ninth rib (C), and craniocaudal dimension of the thoracic spine from the T2–T9 vertebrae (D). (E–G) Measurements of the segmented anatomy in the sagittal projection (E) and 3D rendering (F) are compared with the corresponding anatomy DICOM measurements (G). Measurements made in the 3D rendering must be verified through rotation of the 3D rendering to avoid off-axis results. This can be avoided by obtaining measurements in the appropriate projection (E). lt = left, rt = right.
The accuracy of segmentation is also verified by obtaining measurements of the anatomy at different landmarks on DICOM images and comparing them with segmented masks. As the target anatomy may vary case by case, at our institution, verification measurements for thoracic spine models include (a) the distance from the superior endplate of the uppermost vertebral body to the bottom endplate of the lowermost vertebral body included in the 3D printed model, (b) the anteroposterior dimension of the uppermost vertebra, and (c) the transverse dimension of the widest structure to be included in the 3D printed model, such as the ribs (Fig 11).
Once segmentation is performed and validated, a three-step QA protocol (preprinting, 3D printing, and postprinting) must be performed. Preprinting QA requires validation of the accurate conversion of masks into CAD 3D models. This process involves checking the digital files for any errors, such as holes, gaps, inverted vertices, and intersecting surfaces. The digital file should also be optimized for 3D printing by eliminating nonessential features such as floating islands, reducing the number of triangles, and orienting the model to ensure optimal printing results.
Three-dimensional printing QA requires verification of the 3D printer because it must be calibrated and capable of producing accurate and consistent results. All 3D printers must be regularly inspected, calibrated, and maintained according to the manufacturer’s recommended protocols. Postprinting QA requires inspection of the 3D printed anatomic model and assessment of dimensional accuracy, surface quality, and feature completeness. This can be done by visual inspection and by measuring the model with calipers to compare the 3D printed model with the CAD and DICOM measurements (Fig 12).
Figure 12.
QC after 3D printing ensures that the 3D printed model accurately reflects the segmentation and corresponding ROIs (corresponds to case 3 in Table 1). (A–C) Photographs show 3D printed models that are verified by obtaining measurements at three points: the anteroposterior dimension of the vertebral body (A), the transverse dimension of the ninth rib (B), and the craniocaudal dimension of the spine (C), to ensure print accuracy. (D) Measurements of the 3D printed model (not shown) are compared with the corresponding 3D rendering (bottom; ROI) measurements (top). The results in addition to the type of 3D printer used are documented before model delivery.
Last, the location and format of the segmented data, Standard Tessellation Language (STL) files, and 3D printing files must be documented. Archiving data files used for 3D printing such as STL and OBJ files as a DICOM dataset is feasible but requires participation of information technology leaders and resources (as those data files are usually reserved for the long-term storage of imaging-related patient data) in addition to further vendor support for this feature (35). Additional quality safeguards include documentation of the 3D printer (technology, device specifics, etc), the materials used (expiration date, recall information, etc), and the results of failure mode and effect analysis. By implementing a comprehensive quality management system, medical professionals can ensure that 3D printed anatomic models meet the necessary standards of accuracy, precision, and safety and are suitable for use in surgical planning, medical education, and other similar applications.
Surgical Rehearsal
Evaluation of the 3D CAD model with the operating surgeon allows review of the anatomy of interest as well as at-risk anatomy to aid surgical planning. The surgical planes are prescribed by the operating surgeon based on their surgical approach (posterior, intrathoracic, etc) and ideal margins around the tumor. Ideal surgical margins extend 2 cm beyond the edge of the disease whenever possible. Limitations include insufficient space to accommodate hardware, proximity to joints, and insufficient residual structure(s) compromising stability and functionality. Once the surgeon designates the ideal surgical resection path, analytical primitives including lines, arcs, and datum planes are generated. Planes are edited to remove any intersections or overlapping planes (Movie 6).
Movie 6:
The surgical plan may incorporate the 3D printed model through the addition of surgical planes prescribed by the surgeon. Prescribing digital surgical planes requires verification of accuracy of segmentation, prescription of perimeter planes by the surgeon, trim of plane intersects in preparation to the integration of the planes into the model, inclusion of the final surgical planes within the 3D printed model, and verification of surgical planes by the surgeon.
The analytical primitives are then converted into parts. Boolean operations including union and subtraction may be used to create a singular part that may be integrated into the 3D printed model (Fig 13). The 3D CAD model provides the surgeon an opportunity to prescribe ideal surgical planes that can be incorporated onto the 3D CAD model and 3D printed model, facilitating surgical planning and rehearsal. The addition of these planes onto the model allows the surgeon to verify their surgical plan. Inclusion of the full extent of disease within the area demarcated by the prescribed surgical planes is verified by overlaying the segmentation masks and digital surgical planes onto the registered CT and MR images. Once verified, the resulting 3D CAD model can be 3D printed (Fig 14).
Figure 13.
Surgical planes can be prescribed by the operating surgeon when reviewing the digital model or 3D rendering (corresponds to case 4 in Table 1). (A) CT images show how coronal (top left), axial (top right), and sagittal (bottom left) digital surgical planes can be prescribed in the segmentation software on review of all the parts to be included in the 3D printed model (bottom right). (B) 3D rendering shows digital surgical planes integrated into the 3D model (arrows) that delineate the superior, inferior, and posterior extent of the tumor (red) as prescribed by the operating surgeon. (C) Photograph of the 3D printed model (left) and close-up view (right) show a clear raised textured band (arrows in A–C) integrated within the model representing the surgical planes prescribed at the time of digital rehearsal.
Figure 14.
Verification of 3D printed models based on multimodality examinations requires overlay of the ROI onto the source DICOM dataset (case 3 in Table 1). (A) Sagittal CT image shows ROI overlay, which allows verification of the adequacy of segmentation of the calcified tumor (*) and thoracic spine (white). (B) Sagittal MR image shows ROI overlay, which allows verification of the adequacy of segmentation of the cartilaginous cap and soft-tissue extent (arrows). (C) Sagittal image with overlay of the CT-based ROI (orange) and MRI-based ROI (magenta) shows the total extent of disease. (D, E) Resulting digital CAD 3D model images show the various tumor components, calcified (D) and soft tissue (E), as distinct parts. (F) Photograph shows the 3D printed model that was printed using a Boolean union operation to generate the total tumor (orange), which also includes the ribs and thoracic spine (clear). (G) Photograph shows the 3D printed model used for surgical rehearsal that allows the surgeon to visualize and manually prescribe ideal surgical planes (arrows) onto the 3D printed model. Irregular or discontinuous surgical planes may be difficult to reproduce as CAD-generated surgical planes and integrate them into the 3D printed model before printing. These can be annotated on the model and validated during intraoperative rehearsal by using the surgical navigation software.
Surgical planes may be used as part of the 3D CAD model only, incorporated into the 3D printed model (Fig 13), or used as manual annotations on the 3D printed model (Fig 14). Factors such as the complexity of tumor extent, tumor location, and surgical subspecialties involved in the case influence the way surgical planes are prescribed. As the STL files corresponding to the surgical planes part can be exported into the surgical navigation software, the surgical planes may also be verified using intraoperative navigation software. The surgical planes annotated onto the 3D printed model may be verified using navigation hardware and software (Fig 15).
Figure 15.
Intraoperative surgical navigation system integrates intraoperative imaging and localizers to provide spatial orientation (case 3 in Table 1). (A) Photograph shows surgical rehearsal using the 3D printed model, which depicts the plane (arrowhead) depicted by the navigation probe (arrow). (B) Photograph shows how a framework is used to serve as reference onto the patient (★) and probe (arrow). (C) Photograph shows how the system displays intraoperative images and superimposes the surgical planes (arrowheads) corresponding to the surgical instruments and probe. En bloc resection of the tumor (*) was achieved with the aid of the intraoperative surgical navigation system.
Intraoperative Use of Surgical Planes and Impact of 3D Printed Models
Intraoperative navigation systems are advanced imaging technologies used in surgical procedures to guide the placement of surgical instruments and implants, as well as osteotomes, for precise sectioning. These systems use surface-based registration methods or the iterative closest-point algorithm to align the surface mesh onto the intraoperative images. Intraoperative grids on the patient help with the spatial orientation of the navigation system. These are crucial during two-stage multiday surgery where instrumentation and hardware placement is done in a stepwise fashion (Movies 7, 8). Fiducial markers are also added to the STL files corresponding to the surgical planes to help with manual alignment in case of failure of the navigation system semiautomatic registration. Therefore, these systems can seamlessly integrate the STL files with the intraoperative CT images and depict the spatial orientation of the surgical instruments. Overlay of the tumor STL file is particularly important because intraoperative CT may underestimate the extent of disease as the soft-tissue component is usually better depicted on MR images. By using real-time imaging combined with spatial positioning and orientation, the surgeon can make precise and accurate intraoperative decisions, reducing the need for extended surgical time and minimizing the potential for complications (Fig 15).
Movie 7:
Intraoperative CT clips prior to and following instrumentation demonstrate the phased approach to en bloc removal of the tumor (arrow). Multistep 2-day surgical removal of the bulky tumor using a phased approach to multilevel laminectomies and resection of the posterior elements from the T4-T8 levels and posterior stabilization hardware.
Movie 8:
Cinematic rendering of the pre- and postsurgical CT acquisition demonstrates the chondro-osseous calcifications within the chondrosarcoma (arrow) and corresponding surgical defect following multilevel laminectomies and resection of the posterior elements from the T4-T8 levels and use of posterior stabilization hardware (green rods). Following instrumentation, the alignment and curvature are satisfactory.
Conclusion
Three-dimensional printing technology has the potential to revolutionize complex spine surgery by allowing the creation of patient-specific 3D printed anatomic models and surgical planes. The three-dimensional printing workflow for complex spine tumors often involves integration of multimodality imaging examinations as different technologies and imaging sequences demonstrate various tumor features, influencing our ability to delineate disease extent. Therefore, registration of these examinations and various sequences is essential to achieve accurate segmentation. Verification of the segmented anatomy is crucial to ensure that the full extent of disease is included, in addition to the anatomy to be manipulated at the time of surgery, to provide an optimal surgical rehearsal experience. The choice of materials used for 3D printing influences the visualization of disease involving the epidural space and in cases of intramedullary extension. The use of transparent or clear rigid material for bones against colored tumors better depicts the disease extent. Additionally, using tissue-mimicking materials allows rehearsal of surgical maneuvers. Therefore, using a variety of materials with different tensile strengths is beneficial when creating 3D printed models for complex spine cases.
QA steps are critical in medical 3D printing. QA and QC performed at every step of the workflow yields models that are accurate, precise, safe, and highly reliable. Prescribing surgical planes is an added benefit of surgical rehearsal that can be derived and integrated into 3D printed models. Digital surgical planes verified during surgical rehearsal in conjunction with the intraoperative navigation system may be used for added intraoperative guidance and further optimization of surgical outcomes.
Funding.—The University of Texas MD Anderson Cancer Center is supported in part by the National Institutes of Health through Cancer Center Support Grant P30CA016672.
Presented as an education exhibit at the 2022 RSNA Annual Meeting.
Disclosures of conflicts of interest.—: E.M.A. Owns stock in 3D Systems. All other authors, the editor, and the reviewers have disclosed no relevant relationships.
Abbreviations:
- CAD
- computer-aided design
- DICOM
- Digital Imaging and Communication in Medicine
- QA
- quality assurance
- QC
- quality control
- ROI
- region of interest
- STL
- Standard Tessellation Language
- 3D
- three dimensional
References
- 1. Cobb CA 3rd , Leavens ME , Eckles N . Indications for nonoperative treatment of spinal cord compression due to breast cancer . J Neurosurg 1977. ; 47 ( 5 ): 653 – 658 . [DOI] [PubMed] [Google Scholar]
- 2. Walsh GL, Gokaslan ZL, McCutcheon IE, et al . Anterior approaches to the thoracic spine in patients with cancer: indications and results . Ann Thorac Surg 1997. ; 64 ( 6 ): 1611 – 1618 . [DOI] [PubMed] [Google Scholar]
- 3. Choi D , Bilsky M , Fehlings M , Fisher C , Gokaslan Z . Spine Oncology-Metastatic Spine Tumors . Neurosurgery 2017. ; 80 ( 3S ): S131 – S137 . [DOI] [PubMed] [Google Scholar]
- 4. Choi D, Fox Z, Albert T, et al . Prediction of Quality of Life and Survival After Surgery for Symptomatic Spinal Metastases: A Multicenter Cohort Study to Determine Suitability for Surgical Treatment . Neurosurgery 2015. ; 77 ( 5 ): 698 – 708 . [DOI] [PubMed] [Google Scholar]
- 5. Laufer I, Rubin DG, Lis E, et al . The NOMS framework: approach to the treatment of spinal metastatic tumors . Oncologist 2013. ; 18 ( 6 ): 744 – 751 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Patchell RA, Tibbs PA, Regine WF, et al . Direct decompressive surgical resection in the treatment of spinal cord compression caused by metastatic cancer: a randomised trial . Lancet 2005. ; 366 ( 9486 ): 643 – 648 . [DOI] [PubMed] [Google Scholar]
- 7.Lador R, Regev G, Salame K, Khashan M, Lidar Z. Use of 3-Dimensional Printing Technology in Complex Spine Surgeries. World Neurosurg 2020;133:e327–e341. [DOI] [PubMed] [Google Scholar]
- 8. Parthasarathy J , Jonard B , Rees M , Selvaraj B , Scharschmidt T . Virtual surgical planning and 3D printing in pediatric musculoskeletal oncological resections: a proof-of-concept description . Int J CARS 2023. ; 18 ( 1 ): 95 – 104 . [DOI] [PubMed] [Google Scholar]
- 9. Parr WCH , Burnard JL , Wilson PJ , Mobbs RJ . 3D printed anatomical (bio)models in spine surgery: clinical benefits and value to health care providers . J Spine Surg 2019. ; 5 ( 4 ): 549 – 560 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Morgan C , Khatri C , Hanna SA , Ashrafian H , Sarraf KM . Use of three-dimensional printing in preoperative planning in orthopaedic trauma surgery: A systematic review and meta-analysis . World J Orthop 2020. ; 11 ( 1 ): 57 – 67 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Ravi P , Burch MB , Farahani S , et al. ; University of Cincinnati 3D Printing Clinical Service Participants. Utility and Costs During the Initial Year of 3D Printing in an Academic Hospital . J Am Coll Radiol 2023. ; 20 ( 2 ): 193 – 204 . [DOI] [PubMed] [Google Scholar]
- 12. Krauel L, Fenollosa F, Riaza L, et al . Use of 3D Prototypes for Complex Surgical Oncologic Cases . World J Surg 2016. ; 40 ( 4 ): 889 – 894 . [DOI] [PubMed] [Google Scholar]
- 13. Liew Y , Beveridge E , Demetriades AK , Hughes MA . 3D printing of patient-specific anatomy: A tool to improve patient consent and enhance imaging interpretation by trainees . Br J Neurosurg 2015. ; 29 ( 5 ): 712 – 714 . [DOI] [PubMed] [Google Scholar]
- 14. Santiago L, Volk RJ, Checka CM, et al . Acceptability of 3D-printed breast models and their impact on the decisional conflict of breast cancer patients: A feasibility study . J Surg Oncol 2021. ; 123 ( 5 ): 1206 – 1214 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Hsu MR , Haleem MS , Hsu W . 3D Printing Applications in Minimally Invasive Spine Surgery . Minim Invasive Surg 2018. ; 2018 : 4760769 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Izatt MT, Thorpe PL, Thompson RG, et al . The use of physical biomodelling in complex spinal surgery . Eur Spine J 2007. ; 16 ( 9 ): 1507 – 1518 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kim MP , Ta AH , Ellsworth WA 4th , Marco RA , Gaur P , Miller JS . Three dimensional model for surgical planning in resection of thoracic tumors . Int J Surg Case Rep 2015. ; 16 : 127 – 129 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Li C, Yang M, Xie Y, et al . Application of the polystyrene model made by 3-D printing rapid prototyping technology for operation planning in revision lumbar discectomy . J Orthop Sci 2015. ; 20 ( 3 ): 475 – 480 . [DOI] [PubMed] [Google Scholar]
- 19. Mao K, Wang Y, Xiao S, et al . Clinical application of computer-designed polystyrene models in complex severe spinal deformities: a pilot study . Eur Spine J 2010. ; 19 ( 5 ): 797 – 802 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Mizutani J, Matsubara T, Fukuoka M, et al . Application of full-scale three-dimensional models in patients with rheumatoid cervical spine . Eur Spine J 2008. ; 17 ( 5 ): 644 – 649 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. van Dijk M , Smit TH , Jiya TU , Wuisman PI . Polyurethane real-size models used in planning complex spinal surgery . Spine 2001. ; 26 ( 17 ): 1920 – 1926 . [DOI] [PubMed] [Google Scholar]
- 22. Wu C , Tan L , Lin X , Hu H . Clinical Application of Individualized Reference Model of Sagittal Curves by Three-dimensional Printing Technique and Compuer-aided Navigation System for Lumbar Spondylolisthesis [in Chinese] . Zhongguo Xiu Fu Chong Jian Wai Ke Za Zhi 2015. ; 29 ( 6 ): 734 – 740 . [PubMed] [Google Scholar]
- 23. Xiao JR, Huang WD, Yang XH, et al . En Bloc Resection of Primary Malignant Bone Tumor in the Cervical Spine Based on 3-Dimensional Printing Technology . Orthop Surg 2016. ; 8 ( 2 ): 171 – 178 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Yang M, Li C, Li Y, et al . Application of 3D rapid prototyping technology in posterior corrective surgery for Lenke 1 adolescent idiopathic scoliosis patients . Medicine (Baltimore) 2015. ; 94 ( 8 ): e582 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Zhao W, Shen C, Cai R, et al . Minimally invasive surgery for resection of ossification of the ligamentum flavum in the thoracic spine . Wideochir Inne Tech Malo Inwazyjne 2017. ; 12 ( 1 ): 96 – 105 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Habib A, Jovanovich N, Muthiah N, et al . 3D printing applications in spine surgery: an evidence-based assessment toward personalized patient care . Eur Spine J 2022. ; 31 ( 7 ): 1682 – 1690 . [DOI] [PubMed] [Google Scholar]
- 27. Orguc S , Arkun R . Primary tumors of the spine . Semin Musculoskelet Radiol 2014. ; 18 ( 3 ): 280 – 299 . [DOI] [PubMed] [Google Scholar]
- 28. Ropper AE , Cahill KS , Hanna JW , McCarthy EF , Gokaslan ZL , Chi JH . Primary vertebral tumors: a review of epidemiologic, histological, and imaging findings—Part I:benign tumors . Neurosurgery 2011. ; 69 ( 6 ): 1171 – 1180 . [DOI] [PubMed] [Google Scholar]
- 29. Ropper AE , Cahill KS , Hanna JW , McCarthy EF , Gokaslan ZL , Chi JH . Primary vertebral tumors: a review of epidemiologic, histological and imaging findings—part II: locally aggressive and malignant tumors . Neurosurgery 2012. ; 70 ( 1 ): 211 – 219 ; discussion 219. [DOI] [PubMed] [Google Scholar]
- 30. Philipp MO , Kubin K , Mang T , Hörmann M , Metz VM . Three-dimensional volume rendering of multidetector-row CT data: applicable for emergency radiology . Eur J Radiol 2003. ; 48 ( 1 ): 33 – 38 . [DOI] [PubMed] [Google Scholar]
- 31. Honda O, Johkoh T, Yamamoto S, et al . Comparison of quality of multiplanar reconstructions and direct coronal multidetector CT scans of the lung . AJR Am J Roentgenol 2002. ; 179 ( 4 ): 875 – 879 . [DOI] [PubMed] [Google Scholar]
- 32. Liu Y, Li R, Fan Y, et al . The influence of anisotropic voxel caused by field of view setting on the accuracy of three-dimensional reconstruction of bone geometric models . AIP Adv 2018. ; 8 ( 8 ): 085111 . [Google Scholar]
- 33. Bastawrous S, Wu L, Liacouras PC, et al . Establishing 3D Printing at the Point of Care: Basic Principles and Tools for Success . RadioGraphics 2022. ; 42 ( 2 ): 451 – 468 . [DOI] [PubMed] [Google Scholar]
- 34. Bastawrous S, Wu L, Strzelecki B, et al . Establishing Quality and Safety in Hospital-based 3D Printing Programs: Patient-first Approach . RadioGraphics 2021. ; 41 ( 4 ): 1208 – 1229 . [DOI] [PubMed] [Google Scholar]
- 35. DICOM Standards Committee . Working Group 17. Supplement 205: DICOM Encapsulation of STL Models for 3D Manufacturing . Digital Imaging and Communications in Medicine (DICOM) . https://www.dicomstandard.org/docs/librariesprovider2/dicomdocuments/news/ftsup/docs/sups/sup20502fa41aeacb84aae9b0d9f3bfa0f510a.pdf?sfvrsn=a9571046_2. Published April 3, 2018.