This paper outlines automated processes for nonhuman primate neurosurgical planning based on magnetic resonance imaging (MRI) scans. These techniques use procedural steps in programming and design platforms to support customized implant design for NHPs. The validity of each component can then be confirmed using three-dimensional (3D) printed life-size anatomical models.
This paper describes an in-house method of 3D brain and skull modeling from magnetic resonance imaging (MRI) tailored for nonhuman primate (NHP) neurosurgical planning. This automated, computational software-based technique provides an efficient way of extracting brain and skull features from MRI files as opposed to traditional manual extraction techniques using imaging software. Furthermore, the procedure provides a method for visualizing the brain and craniotomized skull together for intuitive, virtual surgical planning. This generates a drastic reduction in time and resources from those required by past work, which relied on iterative 3D printing. The skull modeling process creates a footprint that is exported into modeling software to design custom-fit cranial chambers and headposts for surgical implantation. Custom-fit surgical implants minimize gaps between the implant and the skull that could introduce complications, including infection or decreased stability. By implementing these pre-surgical steps, surgical and experimental complications are reduced. These techniques can be adapted for other surgical processes, facilitating more efficient and effective experimental planning for researchers and, potentially, neurosurgeons.
Nonhuman primates (NHPs) are invaluable models for translational medical research because they are evolutionarily and behaviorally similar to humans. NHPs have gained particular importance in neural engineering preclinical studies because their brains are highly relevant models of neural function and dysfunction1,2,3,4,5,6,7,8. Some powerful brain stimulation and recording techniques, such as optogenetics, calcium imaging, and others, are best served with direct access to the brain through cranial windows9,10,11,12,13,14,15,16,17,18,19,20,21,22,23. In NHPs, cranial windows are often achieved with a chamber and an artificial dura to protect the brain and support long-term experimentation8,10,12,17,18,24,25,26,27. Likewise, headposts often accompany chambers to stabilize and align the head during experiments14,15,25,26,28,29,30. The effectiveness of these components is heavily dependent on how well they fit into the skull. A closer fit to the skull promotes bone integration and cranial health by decreasing the likelihood of infection, osteonecrosis, and implant instability31. Conventional design methods, such as manually bending the headpost during surgery25,29 and estimating the skull curvature by fitting circles to coronal and sagittal slices of magnetic resonance (MR) scans9,12 can introduce complications due to imprecision. Even the most precise of these create 1-2 mm gaps between the implant and the skull, providing space for granulation tissue to accumulate29. These gaps additionally introduce difficulty placing screws in surgery9, compromising the stability of the implant. Customized implants have more recently been developed to improve osseointegration and implant longevity9,29,30,32. Additional costs have accompanied advancements in custom implant design because of the reliance on computational models. The most accurate methods require sophisticated equipment such as computerized tomography (CT) machines in addition to MR Imaging (MRI) machines30,32,33 and even computer numerical control (CNC) milling machines for developing implant prototypes25,29,32,34. Gaining access to both MRI and CT, particularly for use with NHPs, may not be feasible for labs in need of custom-fitted implants like cranial chambers and headposts.
As a result, there is a need in the community for inexpensive, accurate, and non-invasive techniques of neurosurgical and experimental planning that facilitate the design and validation of implants prior to use. This paper describes a method of generating virtual 3D brain and skull representations from MR data for craniotomy location planning and the design of custom cranial chambers and headposts that fit the skull. This streamlined procedure provides a standardized design that can benefit experimental outcomes and the welfare of the research animals. Only MRI is required for this modeling because both bone and soft tissue are depicted in MRI. Instead of using a CNC milling machine, models can be 3D printed inexpensively, even when multiple iterations are required. This also allows for the final design to be 3D printed in biocompatible metals such as titanium for implantation. Additionally, we describe the fabrication of an artificial dura, which is placed inside the cranial chamber upon implantation. These components can be validated pre-surgically by fitting all parts onto a life-size, 3D-printed model of the skull and brain.
All procedures involving animals were approved by the Institute for Animal Care and Use Committee at the University of Washington. A total of four adult male rhesus macaques (Macaca mulatta) were used in this study. At the time of MRI acquisition, monkey H was 7 years old, monkey L was 6 years old, monkey C was 8.5 years old, and monkey B was 5.5 years old. Monkeys H and L were implanted with custom chronic chambers at 9 years of age.
1. Skull and brain isolation (Figure 1)
2. Craniotomy location planning (Figure 2)
3. Cranial chamber design (Figure 3)
4. Headpost design (Figure 4)
5. Artificial dura fabrication 11 (Figure 5)
6. Fixing holes procedure
These components were previously validated using a combination of MRI visualizations and 3D-printed anatomical models. By comparing the automated craniotomy visualization to the 3D printed craniotomy and the MRI at the location of the craniotomy, it is evident that the virtual craniotomy representation accurately reflects the region of the brain that can be accessed with the specified craniotomy location (Figure 2A–F). Additionally, the accuracy of the automated craniotomy visualization was further evaluated by comparing the virtual representation to existing craniotomies from implantation surgeries (Figure 2E,G). The 3D printed model, automated visualization, MRI, and actual craniotomy highlight the same region, showing the major sulci at the same location and with proportional consistency. The process of brain and skull isolation and subsequent craniotomy visualization takes under 15 min to complete, allowing for several locations to be tested in under 1 h.
The efficacy of the brain isolation procedure was confirmed by comparing the virtual craniotomy to the MRI representation of the craniotomy location (Figure 2B,C,E,F). The similarities indicated that the brain isolation procedure has the capability to represent the correct size, location, and shape of anatomical structures on the brain that are being targeted, such as the sulci.
The combined 3D-printed brain and skull were used as an anatomically accurate model to validate the chamber and headpost designs. Prior to investing in titanium parts, the chamber and headpost were 3D printed in plastic. It was confirmed that the implants fit into the skull and that they were not overlapping with one another or obstructing important anatomical markers. The chamber and headpost design process produced components that matched the curvature of the skull (Figure 3G,I, Figure 4E, Figure 6, Figure 7). The artificial dura was also confirmed to fit adjacent to the inner walls of the chamber with a minor gap to account for adjustments made during implantation. Custom chambers were implanted in two macaques. Contrary to previous chamber design methods9, every screw that was attempted to be inserted was able to be screwed in. This is due to the drastic reduction of gaps between the chamber and the skull with the custom fit in comparison to the chamber designed from MRI curvature approximations9 (Figure 6A–F). One custom-fit chamber has been implanted for over 2 years, and the other a year and a half. With proper maintenance, there has been no screw loss, infection, or stability issues that have arisen due to these implants (Figure 3I).
The custom headpost and chamber design processes prevent the need for manual adjustments during surgery, which could otherwise add hours to the surgery duration. These techniques also decrease the 1-2 mm gaps that result from curvature approximations29, fostering better implant health and improving experimental outcomes. The refinements prevent complications with the implant and extend implant longevity, therefore also improving animal welfare.
Figure 1: Brain and skull isolation. (A) Layered magnetic resonance image (MRI) coronal slices. (B) Layered binary mask from skull thresholding. (C) Layered slices of the isolated skull from an inverted binary mask. (D) Reconstructed 3D skull. (E) Layered binary mask from brain thresholding. (F) Layered MRI slices of isolated brain. (G) Reconstructed 3D brain. Please click here to view a larger version of this figure.
Figure 2: Craniotomy planning. (A) Craniotomy visualization with 3D printed brain and skull model for Monkey B. (B) Craniotomy visualization in computational software for Monkey B. (C) Craniotomy visualization in magnetic resonance (MR) image for Monkey B. (D) Craniotomy visualization with 3D printed brain and skull model for Monkey H. (E) Craniotomy visualization in computational software for Monkey H. (F) Craniotomy visualization in Magnetic Resonance (MR) image for Monkey H. (G) Image of craniotomy in Monkey H. Please click here to view a larger version of this figure.
Figure 3: Chamber implant design. (A) Skull region (gray) used for STL resolution reduction. (B) Skull STL resolution reduction in SOLIDWORKS. (C) Chamber inner ring, highlighted. (D) Chamber Skirt Design in SOLIDWORKS. (E) Connecting chamber skirt and top. (F) Chamber STL in SOLIDWORKS. (G) 3D printed brain, skull, and chamber. (H) Titanium chamber. (I) Implanted chamber in Monkey H. Please click here to view a larger version of this figure.
Figure 4: Headpost design. (A) Headpost bottom outline on skull STL resolution reduction. (B) Custom-fit headpost footprint. (C) Headpost bottom. (D) Headpost design in SOLIDWORKS. (E) 3D printed headpost on the skull. (F) Titanium headpost. Please click here to view a larger version of this figure.
Figure 5: Artificial dura fabrication. (A) Clamping of silicone mixture using mold. (B) Artificial Dura. This figure has been adapted with permission from Griggs et al.11. Please click here to view a larger version of this figure.
Figure 6: Custom-fit versus skull curvature fit chamber. Chamber designed from MRI curvature estimations on skull9 from an (A) anterior view, (B) side view, and (C) posterior view. Custom designed chamber from a (D) anterior view, (E) side view, and (F) posterior view. Please click here to view a larger version of this figure.
Figure 7: Chamber, headpost, and artificial dura on overlaid brain and skull Please click here to view a larger version of this figure.
Supplementary Figure 1: Thresholding and craniotomy location planning. (A) Example binary mask with a suitable threshold. (B) Coronal slice on MRI for identifying craniotomy location. Please click here to download this file.
Supplementary Figure 2: Process of STL File Reduction in MATLAB for the chamber design. Please click here to download this file.
Supplementary Figure 3: Visual representation of a hole in the skull STL resolution reduction. Please click here to download this file.
Supplementary Figure 4: Chamber skirt software screenshots. (A) Inner ring of the chamber skirt and the inner surface of the chamber top as concentric mates. (B) Translating chamber skirt downwards. Please click here to download this file.
Supplementary Figure 5: Chamber skirt and chamber top with and without overlap. (A) Under-view example of overlap between the chamber skirt and the chamber top (Modifies the lower surface of the chamber skirt). (B) Example of no overlap between chamber skirt and chamber top. Please click here to download this file.
Supplementary Figure 6: Planes obstructing screw holes and elimination of obstruction. (A) Example of planes obstructing the screw holes following screw hole placement. (B) Outline of extruded cut to eliminate surfaces inside of screw holes. Please click here to download this file.
Supplementary Figure 7: Point selection and the axial plane of the skull. (A)Point selection for headpost design. (B) Upper view of the plane parallel to the axial plane of the skull. (C) Side view of the plane parallel to the axial plane of the skull. Please click here to download this file.
Supplementary Figure 8: Example of mates. (A) First mate – Top surface of the circular headpost platform and the bottom surface of the headpost top as concentric mates. (B) Second mate – Edge of the top surface of circular headpost platform and edge of the bottom surface of the headpost top as concentric mates. (C) Third mate – A line going vertically along the back leg of the headpost and a line running horizontally along the back of the headpost top as perpendicular mates. Please click here to download this file.
Supplementary Figure 9: Fixing holes procedure. (A) Knitted surfaces surrounding the gap in the imported surface. (B) Axis on each point at the edge of the knitted surface. (C) End result of fixing holes procedure. Please click here to download this file.
Supplementary Figure 10: Performing extruded cut. (A) Extruded cut surrounding extrusions from fixing holes procedure. (B) Example extruded cut to a plane on the top surface of the chamber bottom. Please click here to download this file.
Supplemental Coding File 1: Coding files for the protocol. Please click here to download this file.
This paper outlines a straightforward and precise method of neurosurgical planning that is not only beneficial for the development of components used for NHP cranial window implantation but also transferrable to other areas of NHP neuroscience research13,15,25. In comparison to other current methods of NHP implant planning and design25,29,30, this procedure has the potential to be adopted by more neuroscience labs because it is simple and economical. While CT is commonly used for skull modeling32,38, this protocol provides sufficient modeling detail for both the brain and the skull using only MRI scans. Existing methods require both MRI and CT scans for brain and skull isolation30,32,33, while this method eliminates additional costs and challenges of CT imaging. An additional benefit is that this model does not require the alignment of MRI and CT scans, saving significant time and preventing issues associated with poor alignment39. Generating both brain and skull models from a single imaging file produces highly compatible models easily combined for craniotomy visualization. This feature is particularly useful for iterative craniotomy testing processes because rather than combining and aligning files from separate programs30,33, both models are generated in one software from a single input file and display automatically within seconds. This allows for efficient confirmation of brain and skull modeling accuracy and ensures implants will match the skull curvature in vivo. This also eliminates iterative 3D printing of the skull previously required for determining the optimal craniotomy location35, thus saving tens of hours of printing per iteration. Our software-based technique, by comparison, takes around 10-15 min to generate each craniotomy iteration.
Identifying the implant location relative to the frontal, parietal, and temporal skull regions, as well as other skull features, has immense benefits for surgical and experimental planning. This feature is capitalized on to custom design the headpost footprint with respect to the chamber footprint. For any NHP neuroscience research, this spatial modeling feature can be adapted to design components from anatomical planes, MRI coordinates, anatomical features of the brain and skull, and with respect to existing implants. By doing this, the possibility of unforeseen issues during or after implantation is drastically reduced. This procedure also has the ability to create implants that span multiple brain areas from different planes while maintaining a tight fit to the skull.
The method highlighted here creates a circular chamber and allows for a headpost to be designed around the chamber. However, the procedure here has the potential to accommodate other shapes through the modification of the Chamber Skirt Design section. The same is true for the headpost design – the procedure allows for different numbers of legs and other custom shapes to be created, with the shape being primarily dependent on the available space around the chamber. The shape of the skull STL reduction, which is currently a ring for the chamber design, could be further modified to create different skull STL reduction shapes tailored to the need of particular chamber or headpost designs, facilitating more efficient adaptation.
Although this process effectively creates customized implants, there are steps that can be improved upon for more efficient production. As mentioned before, aligning the top of the headpost perpendicular to the skull is an iterative process with the outlined method in this paper due to the difficulty of identifying the skull orientation in the design software. To streamline the process of positioning the headpost top on the bottom part, additional markers could be placed on the virtual skull representation to indicate axial, sagittal, and coronal planes. The protocol also has the potential to be further automated for increased ease of use. While the skull STL reduction method discussed in this protocol is effective for designing implants, it could be made faster and more consistent with further automation. Our validation procedure requires 3D printing of the skull and implant prototypes for verification that the implants matched the curvature of the skull. This step could potentially be eliminated by creating a method of virtual 3D visualization that combines the brain, skull, chamber, headpost, and artificial dura together.
Our platform provides an entirely virtual process of craniotomy planning and custom implant design. The final designs can be 3D printed and verified on a life-size physical model35. Contrary to existing methods, our protocol does not require costly product iterations or access to expensive machinery like CNC milling machines29,34. Similar to other existing methods of implant design9,12,29,30,32,33,40, this method completely relies on an imaging modality to accurately depict anatomical structures. Any inaccuracy present in the MRI scan or changes in brain or skull anatomy between MRI and surgery may compromise the efficacy of the implant. Therefore, proper planning for MRI acquisition is essential to optimizing implant design.
The authors have nothing to disclose.
We would like to thank Toni Haun, Keith Vogel and Shawn Fisher for their technical help and support. This work was supported by the University of Washington Mary Gates Endowment (R.I.), National Institute of Health NIH 5R01NS116464 (T.B., A.Y.), NIH R01 NS119395 (D.J.G., A.Y), the Washington National Primate Research Center (WaNPRC, NIH P51 OD010425, U42 OD011123), the Center for Neurotechnology (EEC-1028725, Z.A., D.J.G.) and Weill Neurohub (Z. I.).
3D Printing Software (Simplify 3D) (Paid) | Simplify3D | Version 4.1 | Used for 3D printing using MakerGear printer |
C-Clamp | Bessey | CM22 | Used for artificial dura fabrication, 2-1/2 Inch Capacity, 1-3/8 Inch Throat |
Formlabs Form 3+ 3D Printer | Formlabs | Form 3+ | Used for precise 3D printing |
MakerGear M2 3D Printer | MakerGear | M2 revG | Used for 3D printing implant prototypes |
MATLAB (Paid) | MathWorks | R2021b | Used for brain and skull isolation, virtual craniotomy visualization and skull STL reduction |
Phillips Acheiva MRI System | Philips | 4522 991 19391 | Used for non-human primate imaging |
Photopolymer Resin | Formlabs | FLGPGR04 | 1L, Grey, used for precise 3D prints with Formlabs printer |
PreForm Print Preparation Software | Formlabs | Version 2.17.0 | Used for 3D printing with Formlabs printer |
Printing Filament (PLA) | MatterHackers | 88331 | PLA 1.75 mm White. Used for 3D printing with MakerGear printer |
Silicone CAT-1300 | Shin-Etsu | Used for artificial dura fabrication | |
Silicone KE1300-T | Shin-Etsu | Used for artificial dura fabrication | |
SolidWorks (Paid) | Dassault Systems | 2020 | Used for chamber and headpost design |
Syn.Flex-S Multicoil | Philips | 45221318123 | Used for non-human primate imaging |