Development and Validation of a 3-Dimensionally Printed Craniotomy Simulation Model for Under-resourced Healthcare Systems: A Technical Note

Abstract

BACKGROUND AND OBJECTIVES:

Cadaver-based simulations are limited by financial and legal constraints, particularly in low- and middle-income countries (LMICs). Three-dimensional (3D) printing–based alternatives are often limited in scope and prohibitively expensive. This report describes a 3D-printed model that addresses these concerns by simulating a variety of craniotomy approaches in a cost-effective manner.

METHODS:

A craniotomy model was developed using medical imaging, 3D printing, and foam molding. Detailed manufacturing information for the model is provided. The model was assessed by 40 neurosurgery residents and young attendings in a Latin American LMIC with a questionnaire administered after they performed simulated craniotomies using the model.

RESULTS:

The total cost of the model was 47.52 USD, and it took 6 hours to produce. Thirty-four (85%) participants agreed that the simulated tissue elastance allowed for using the model to teach surgical approaches. Twenty-six (65%) participants agreed that the model provided realistic haptic feedback. All participants felt the model was useful in learning how to perform emergent craniotomies and that training with the model was transferable to clinical practice. Thirty-seven (93%) participants felt use of the model would help prevent serious intraoperative complications. Thirty-six (90%) participants agreed that the model would be useful in educating patients and families about their neurosurgical problem and the planned surgical approach.

CONCLUSION:

Most participants found that the craniotomy model provided clinically translatable neurosurgical skills, and it may be a viable alternative to cadaver-based training in LMICs.

ABBREVIATION:

LMICs

low- and middle-income countries.

Developing the unique technical skills required to perform neurosurgical procedures requires years of dedicated training involving surgical simulation. Owing to their high-fidelity representation of human tissue, cadavers have long been the gold standard for such simulation.¹ However, several factors limit their use, particularly in low- and middle-income countries (LMICs).^2,3 Cadavers have a high purchase price and are usable only for a limited time, which necessitates frequent cadaver purchases by surgical training programs.² The embalming agents involved with the use of cadavers also present health and environmental risks because some embalming agents have been identified as biohazards and carcinogens.² In addition, the use of cadavers is culturally and legally prohibited in some parts of the world.² The necessity of anatomic training for neurosurgery trainees and the limited availability of scalable teaching models necessitate a high-fidelity alternative.

The advent of affordable desktop 3-dimensional (3D) printing technology has created an opportunity to develop alternatives to cadaver-based surgical simulations that address these challenges. Several models have been created that simulate a diverse array of neurosurgical procedures. A recent review by Blohm et al⁴ identified 40 articles describing 3D-printed models that have been developed to simulate aneurysm, spine, craniosynostosis, transsphenoidal, craniotomy, skull base, and tumor surgeries. Despite the breadth of available models, few have been published with detailed cost information. In addition, several of the articles that provide cost information feature models that would be prohibitively expensive for use in LMICs. For example, Lin et al developed a tuberculum sellae meningioma resection model that cost 700 USD, and Tai et al developed a 1000 USD external ventricular drain placement simulator model.^5,6 Although these models may include features that are not possible in less expensive models, there exists an unmet need for models that simulate common neurosurgical procedures at a price that is acceptable for institutions in LMICs.

To our knowledge, this technical note describes the only affordable 3D-printed craniotomy simulation model targeted for use in LMICs. Craniotomy was chosen as the focus for this model because it is an essential step in several neurosurgical procedures. Several neurosurgical residents and attending neurosurgeons in an LMIC evaluated the fidelity and educational value of the model. In addition to presenting participant feedback from these simulations, we present detailed production and shipping information for the model.

METHODS

The study protocol was reviewed and approved by the institutional review board of St. Joseph's Hospital and Medical Center in Phoenix, Arizona, and participant consent was obtained. The Thurston Innovation Center at Barrow Neurological Institute is a rapid prototyping laboratory where neurosurgical residents and faculty collaborate with academia and industry on medical device and education projects. Its resources were used to design and manufacture the model described in this technical note. The model was created using MRI data and advanced volumetric modeling from a patient under express consent and institutional review board coverage (PHX-20-500-113-72-08).

Volumetric Modeling

We describe the volumetric modeling and manufacturing processes in the Figure. First, we segmented cortical brain tissue in the initial MRI using brain imaging software (Freesurfer; Martinos Center for Biomedical Imaging). Skull, nerve, eye, blood vessel, and cranial nerve tissue were subsequently segmented using a free open-source extensible software application for medical image computing and visualization (3DSlicer; http://www.slicer.org).⁷ We used a digital sculpting application (ZBrush; Pixologic, Inc.) to smooth the resulting meshes. We then used a 3D computer graphics application (Autodesk Maya; Autodesk, Inc.) to rebuild an .STL model of the vasculature using the segmented vasculature .OBJ file. Next, 3D modeling software (Autodesk Meshmixer; Autodesk, Inc.) was used to refine the meshes and merge mesh models for manufacturing. We converted the resulting file into tool paths for the 3D printers using 3D printing software (Simplify3D [Simplify3D, LLC] and Preform [Formlabs, Inc.]).

FIGURE.

Barrow Neurological Institute neurosurgical training model volumetric modeling and manufacturing process. A, Freesurfer (Martinos Center for Biomedical Imaging) was used to segment brain tissue and 3DSlicer (http://www.slicer.org) was used to segment skull, eye, blood vessel, and nerve tissue based on MRI. ZBrush (Pixologic, Inc.) was used to smooth the resulting meshes. Autodesk Maya (Autodesk, Inc.) was used to build an .STL file of the intracranial vasculature. B, Autodesk Meshmixer (Autodesk, Inc.) was used to further refine the meshes and merge the modeled anatomic structures. Simplify3D (Simplify3D, LLC) and Preform (Formlabs, Inc.) were used to convert the resulting mesh files into 3D printer tool paths. The brain and skull were printed using fusion deposition modeling, and stereolithography was used to print nerve tissue. A silicone mold was created using the 3D-printed model. The final brain tissue model was molded in soft foam. Finally, the skull and nerves were painted, the brain model was inserted into the skull, and the skull hemispheres were fused. Used with permission from Barrow Neurological Institute, Phoenix, Arizona. 3D, 3-dimensional.

Manufacturing

Two 3D printing modalities were used to construct the model. Fused deposition modeling with a 3D printer (Creator Pro; Zhejiang Flashforge 3D Technology Co., Ltd.) was used for the cortical brain tissue, skull, and vasculature. Stereolithography with a 3D printer (Form 2; Formlabs, Inc.) was used to construct the cranial nerves. The cortical brain model created with fusion deposition modeling was used along with a silicone mold to create a final cortical brain tissue model constructed from soft foam. The component costs and labor hours for each of the model components are reported in Table 1. In addition to the costs to manufacture the model, there were shipping costs associated with sending the models to the test institution in Peru. Twenty complete models were shipped in 2 batches at a total cost of 1925.82 USD. Both participating institutions needed to send letters describing the models to the US Transportation Security Administration before they could be shipped.

TABLE 1.

3D-Printed Craniotomy Simulation Model Component Costs and Labor Time

Component	Material cost (USD)	Equipment cost (USD)	Labor (h)
Skull	23.41	8.05	3.00
Brain	4.71	2.80	1.25
Nerves	1.71	5.85	0.50
Assembly	0.99	0.00	1.25
Total	30.82	16.70	6.00

Evaluation

Study inclusion criteria were internet access and being either a neurosurgery resident or neurosurgery attending physician with less than 5 years of experience from a Latin American LMIC. Participants were specifically recruited from a region with established neurosurgical training programs to evaluate the model's potential as a supplement to existing educational resources rather than as a primary training method. We recruited 40 participants using official announcements on the Sociedad Peruana de Neurocirugia (Peruvian Neurosurgical Society) website and social media accounts as well as an email to all Andean Latin American neurosurgical residents in May 2022. These participants all performed craniotomies using the model and completed a questionnaire regarding its anatomic fidelity and educational value. We performed data aggregation and analyses with an open-source program for statistical analysis (JASP; University of Amsterdam). This study followed Strengthening the Reporting of Observational Studies in Epidemiology guidelines.

RESULTS

Questionnaire Results

Results from the questionnaire on the physical model are reported in Table 2. Several questions evaluated the mechanical and anatomic characteristics of the physical model. Most (34 of 40; 85%) of the participants felt the tissue elastance of the brain and cerebellum adequately allowed for teaching surgical approaches. All the participants who disagreed with this statement were attending physicians, and all nonetheless felt that the model was useful for practicing craniotomies and emergent procedures. Twenty-six (65%) of the participants felt that the model provided realistic haptic feedback. Of the 14 participants who felt that the model was not realistic, 12 (86%) attributed this to the brain material. Most participants (38 of 40; 95%) felt that the model had sufficient overall anatomic fidelity. The 2 participants who disagreed with this statement were attending physicians, and they highlighted the lack of cerebral vessel detail and simulated meninges.

TABLE 2.

3D-Printed Craniotomy Simulation Model Questionnaire Responses

Questionnaire item	No. (%) of responses (n = 40)
	Yes	No
The tissue elastance of the brain and cerebellum allowed for teaching surgical approaches.	34 (85)	6 (15)
The model provided realistic haptic feedback.	26 (65)	14 (35)
The model has high anatomic fidelity for craniotomy simulation.	38 (95)	2 (5)
The model is optimized to simulate supratentorial craniotomies.	36 (90)	4 (10)
The model is optimized to simulate infratentorial craniotomies.	30 (75)	10 (25)
In general, the model is useful for learning to perform emergent craniotomies.	40 (100)	0 (0)
The model is useful for learning to prevent surgical disasters such as sinus tears.	37 (93)	3 (8)
The model could be used to educate patients and families about their neurosurgical problem and the planned surgical approach.	36 (90)	4 (10)
Training with the model is transferable to clinical practice.	40 (100)	0 (0)

Additional questions in the survey evaluated the physical model's ability to simulate various craniotomy approaches. Thirty-six participants (90%) felt the model was optimal to learn supratentorial craniotomies, whereas 30 (75%) felt the model was optimal to learn infratentorial craniotomies. Of the 10 participants who felt that the model was not optimal for practicing infratentorial approaches, 8 felt that this was the case because of the lack of upper cervical spine and vertebral arteries. All participants felt that the 3D-printed model was useful in learning how to perform emergent craniotomies in general. In addition, most (37 of 40; 93%) felt that the model was useful for training on how to prevent surgical disasters, such as a sinus tear. Most (36 of 40; 90%) also felt that the model could be used to educate patients and families about their neurosurgical problem and the planned surgical approach. The 4 (10%) participants who disagreed were junior residents.

DISCUSSION

To our knowledge, this article is the first detailed technical note describing the development of an affordable, high-fidelity craniotomy simulation model evaluated for use in LMICs. Cadavers have long been a mainstay of neurosurgical simulation. However, limitations such as their high cost and the short period in which they are usable make them unsuitable for use in LMICs. Advancements in 3D printing technology have enabled several groups to develop neurosurgical simulation models that partially address these limitations, but such models are often too expensive for use in LMICs. Moreover, our model offers significant cost advantages for repeated training sessions. At $47.52 per unit, the model can be used repeatedly over multiple years, whereas cadaveric specimens require ongoing procurement and disposal costs with limited usable periods. The model described in this technical note offers significant realism and cost advantages relative to these previously developed models, which makes it more suitable for use in LMICs. Although this model could benefit neurosurgical training programs globally, we specifically targeted those in LMICs because these programs face the greatest barriers to accessing traditional cadaveric specimens because of cost, legal restrictions, cultural considerations, and infrastructure limitations. Finally, although 3D printing infrastructure varies across LMICs, this barrier is becoming less relevant as desktop 3D printing technology becomes increasingly affordable and accessible.

Synthetic cranial models may represent a sufficiently realistic substitute for use in surgical training in LMICs while mitigating many of the obstacles associated with cadaver use. The model we developed was received favorably by the participants who used it, with all of them agreeing it was useful in learning how to perform emergent craniotomies and that this training was translatable to clinical practice. The realism of the model was similarly viewed favorably, with 38 (95%) agreeing that the model had sufficient overall anatomic fidelity. Mery et al⁸ developed a 3D-printed craniotomy and aneurysm clipping training model that was evaluated by 32 neurosurgery residents. On a 5-point Likert score–based questionnaire, the statement that the material of the model represented the surgical reality in an acceptable way received a mean score of 3.97.⁸ The statement that the model allowed for a craniotomy to be performed properly received a mean score of 4.22.⁸ Although this reusable simulator received favorable ratings, it had an initial cost of 2500 USD and cost 180 USD for subsequent uses, far exceeding the cost of our model. An important distinction must be made regarding the intended role of our simulation model. This model should be viewed as a complement to, rather than a replacement for, comprehensive neuroanatomic education using cadaveric specimens. Our model was specifically designed to simulate the technical aspects of craniotomy procedures, particularly the bone work, initial exposure, and instrument handling, rather than to provide detailed neuroanatomic education. Although cadaveric specimens remain superior for comprehensive anatomic study, our model offers advantages for repetitive technical skill training. In many LMIC settings, the choice is often between a model like ours and no hands-on craniotomy training at all because of resource constraints.

A significant factor limiting the applicability of previously developed models is the high model production cost. However, some affordable models have been developed. Bishop et al⁹ developed an affordable craniotomy and burr simulator model aiming to train general surgeons to perform these procedures in remote and rural areas of Canada where neurosurgeons are not available. The model was evaluated by 16 rural general practitioners, and feedback was generally positive on an open-ended survey completed by 9 of the participants. Although the model's production cost of 44.02 CAD is comparable with that for the model described in this technical note (47.52 USD), it features only a portion of the skull and would not be suitable for simulating some types of craniotomy, such as retrosigmoid craniotomy. Furthermore, it does not feature cranial nerves and vasculature, and it was evaluated by a relatively small cohort of residents.

Our model addresses the limitations of previously developed models in a manner that makes it ideal for use in neurosurgical training in LMICs. The model is capable of simulating all craniotomy types at a cost that is significantly below that of previously developed models. Furthermore, its educational value was assessed by a comparatively large cohort of neurosurgical residents and young attending physicians than previous models. To our knowledge, this is the first model exclusively evaluated by participants in an LMIC. This technical note also provides a comprehensive overview of each step of the manufacturing process and detailed cost information. Future applications could include patient-specific models generated using individual patient imaging data, providing case-specific rehearsal opportunities and enhanced patient counseling tools. The modular design allows for adaptation to simulate specific pathologies by incorporating 3D-printed pathological anatomy.

Limitations

Although the model was initially developed and manufactured in the United States, its long-term viability in LMICs depends on local production. In our deployment, shipping costs (USD 1925.82) were more than twice the total manufacturing cost of all 20 models (USD 950.40), highlighting the cost-prohibitive nature of transcontinental transport. Local fabrication not only eliminates this burden but also reduces per-unit costs and strengthens regional technical capacity. To that end, we are establishing partnerships with university-based engineering programs in Latin America to develop decentralized manufacturing pipelines. Although the model was evaluated in a single LMIC context, further trials across diverse settings are needed to validate generalizability and guide broader implementation.

CONCLUSION

To our knowledge, the craniotomy simulation model described in this technical note is the only one targeted for use in LMICs. The feedback from neurosurgery residents and attending physicians who performed a craniotomy using the model demonstrated that it may be a viable alternative to cadaver-based craniotomy training in LMICs. It also highlighted the applicability of 3D printing to the production of such models and the need for future studies validating this model in additional LMIC environments.

Funding

This study did not receive any funding or financial support.

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

COMMENTS

In the current neurosurgical training environment, the integration of technological advancements and AI tools has become not only inevitable but essential for enhancing teaching and learning. However, a striking disparity remains in the availability and accessibility of such resources between high-income countries (HICs) and low and middle-income countries (LMICs), driven by multiple systemic factors. It is imperative that we strive to bridge this gap to ensure equitable training opportunities for neurosurgical trainees worldwide.

Training tools must be affordable, accessible, and freely available across the globe. In this context, the development and use of 3D-printed craniotomy simulation models represent a significant step forward—particularly for under-resourced healthcare systems. Their potential impact on neurosurgical education in LMICs is both promising and undeniable. I commend the authors for their valuable contribution to neurosurgical literature.

As part of a “frugal” neurosurgical training initiative under our charitable platform, NeusMent (www.neusment.org), we explored the use of coconuts as cranial models for medical students and junior trainees to practice basic cranial procedures. The feedback was overwhelmingly positive, and we successfully replicated this low-cost, innovative training method across several medical schools in the UK.

Simple, eco-friendly, and affordable simulation tools like these hold immense potential for broader application, especially in LMICs. They embody the spirit of resourceful innovation and can play a pivotal role in democratizing neurosurgical training globally.

Chandrasekaran Kaliaperumal

Edinburgh, UK