Neuroendovascular Simulation and Replication

Henry Woo, MD
Chandramouli Sadasivan, PhD
B. Baruch Leiber, PhD

The model of “See one, do one, teach one” in medical training, and in particular, neurosurgery, is now obsolete. Advances in technology and the complexity of pathologies that can be treated effectively transformed the training of medical students, residents, and even experienced practitioners. Furthermore, the evolving economics of health care focus on outcomes and quality. Inexperience resulting in unsatisfactory outcomes are fast becoming unacceptable—especially for complex procedures where errors can result in devastating, life-altering consequences. In neurosurgery, there is a real need to practice specific tasks to avoid the collateral damage to patients when practitioners are in the early stages of the “learning curve.”

The concept of the learning curve was introduced by Hermann Ebbinghaus, a German psychologist in the late 1880s, and shortly thereafter was utilized by Theodore Wright to describe its effects on production costs in the airline industry.1 In addition to the learning curve, the concept of “mastery” and the 10,000-hour rule has been popularized more recently by Malcolm Gladwell in his book Outliers.2 The principle holds that 10,000 hours of practice are needed to become worldclass in any field.

However, the 10,000-hours concept can further be traced back to a 1993 paper “The Role of Deliberate Practice in the Acquisition of Expert Performance” by K. Anders Ericsson.3 Although the paper studied violinists at the Hochschule der Kuenste, the Music Academy of West Berlin, the idea of deliberate practice clearly correlates to the performance of surgical procedures. While the 10,000-hour rule is time-based, deliberate practice involves other conditions that must be met in order to improve performance. The most important are 1) motivation and concentrated effort by the practitioner, 2) feedback on their performance, and 3) progressively increasing difficulty of the tasks being practiced. Surgical procedures, especially in the practice of neurosurgery, are particularly well-suited to satisfy these conditions. First, most neurosurgeons are self-motivated enough to want to improve their technical skills, as it affects patient outcomes as well as clinical productivity and income. Putting the aura of being a brain surgeon in the lay public’s mind (and the decision-making) aside, the performance of the surgery itself, no matter how complex, is still fundamentally manual labor.

Second, neurosurgeons receive quick and real feedback on their performance. If they did not perform a procedure well, there are radiographic and clinical consequences that manifestly reflect that performance. In neurosurgery, the clinical consequence may be death or significant morbidity secondary to neurologic injury. Even if there are no neurologic sequelae to the poorly performed procedure, there are radiographic results, and, as one of my colleagues likes to say: “That is a picture you would be embarrassed to show to your grandmother.”

Third, there is a wide variety in the difficulty of cases that are 1) disease specific, i.e., a lumbar disc versus an odontoid screw versus a tumor or AVM resection, and 2) patient specific for anatomic reasons, i.e., a straightforward narrow-necked posterior communicating aneurysm versus a giant communicating segment aneurysm involving a large portion of the supraclinoid internal carotid artery. Prior to the era of modern simulation, and still existent today, technical mastery requiring those 10,000 hours was obtained in the operating room with the collateral damage being the death or disability those patients suffered.

Medical simulation traces its history back to Galen, Vesalius, and Da Vinci4, but more recently, as technology has changed, the clinical practice of medicine has also changed the applicability of simulation. Animal models of simulation such as sidewall and bifurcation aneurysm models for endovascular coiling, typically do not reflect the complexity associated with actual clinical procedures. The aneurysms are relatively uniform in size and configuration, and there is no tortuosity in the proximal vasculature, which has a profound effect on the performance of endovascular devices. Both of these factors are patient-specific and cannot be easily simulated with animal models. Virtual software-based simulators have also advanced tremendously, but fundamentally they are still highly glorified video games.

The model of virtual simulation has been in the airline industry with flight simulators. In the film Sully, after pilot Chesley Sullenberger landed his debilitated plane on the Hudson River, he was asked in a briefing, “How did you know what to do?” While Sully had never landed a plane with malfunctioning engines on the water prior to this event, he credited his time on flight simulators practicing different disaster/failure scenarios as a major reason why he was able to save those passengers, his crew, and himself. The two major differences between flight simulation and surgical procedure simulation are user interface and haptic feedback. While there are some differences in the exact style of the controls for different parts of planes, e.g., the throttle, the yoke, and rudder pedals, what they control is relatively standardized. Second, haptic feedback is not required for the technical performance of flying modern planes. If the plane is jerking violently or the plane crashes, the mistakes were made much earlier, and the pilot is making decisions not based on feedback from the controls, but from data in the cockpit. For surgical procedures, haptic feedback is absolutely critical to the technical performance of the procedure. How hard you retract on the brain, the spinal cord, or the carotid artery with your suction device or bipolar is critical to your performance as a surgeon. For this reason, there is no virtual simulator that has adequately recreated the neurosurgical environment of removing a brain tumor, placing a pedicle screw, or dissecting a sylvian fissure and clipping an aneurysm.

Virtual simulators in neurosurgery have been most prevalent in endoscopic procedures and endovascular neurosurgery. For endoscopic procedures, haptic feedback, while still important, is not critical as it is for traditional open procedures. Haptic feedback for endovascular procedures, however, is essential. The feedback gives you information about the stability of your guide catheter platform, the likelihood of dissection or perforation of an aneurysm, rupture of a catheter, etc. It is realtime feedback with significant consequences for the final outcome of the procedure. For this reason, virtual simulators in endovascular neurosurgery really do nothing more than give a novice a better understanding of the fundamental steps of a diagnostic angiogram or straightforward coiling. For patient and device specific procedures, it is impossible to model the performance of the procedure virtually. For example, the exact same aneurysm can be coiled thousands of times with the exact same coils, the exact same guide catheter, and microcatheter, etc., but if you examine how the coils were distributed in the aneurysm, the pattern of that distribution would be different in all of them, making it impossible to virtually model. Furthermore, as new technology is developed, there is no prior behavior of the device allowing you to model it virtually. You would need to rely on a software engineer’s guess as to how that new device will behave, which is certainly inferior to the guess of an experienced interventionalist, even if he or she has never used that device before. In the opinion of the authors, virtual simulation, while improving significantly, is still far from providing clinical applicability except for the most basic of procedures. Complex neurosurgery will need to find a different method of simulation if it is going to provide significant value to the experienced practitioner.

Three-dimensional printing has been a revolution across numerous industries and clearly has already significantly impacted our day-to-day lives even beyond medical applications. Three-dimensional printing was invented by Chuck Hull in 1983, when he realized curing photopolymers with light while he was finishing table tops had a potential beyond that relatively straightforward application. He eventually founded 3D Systems (Rock Hill, South Carolina)5. Fused deposition modeling, another form of three-dimensional printing, was invented by S. Scott Crump while he was making toys for his children. Crump eventually founded Stratasys Ltd. (Eden Prairie, Minnesota)5. Prior to the era of stereolithography, rapid prototyping, also known as three-dimensional printing (all are essentially synonymous terms), the creation of three-dimensional models was predominantly performed through a process of investment casting where an outer shell was machined in metal and filled with a material that was malleable, typically a liquid that eventually hardened, resulting in the three-dimensional model. This process was labor- and time-intensive, usually requiring weeks, if not months, to create a single model. Once the outer shell of the investment cast was created, simply duplicating the same model was straightforward, but any change in the anatomy of the model itself required creation of a new cast which meant essentially starting from scratch.

Figure 1. Three-dimensional medical imaging of a patient’s carotid cavernous aneurysm (A) is pruned and converted to stereolithography file format (B) and 3D-printed to obtain the anatomical model (C). The model can be dip-coated with silicone to obtain a replica of the pathology (D) for neuroendovascular simulation.

As medical imaging has advanced with high-resolution MRI, CT, and cone beam CT, it is now possible to convert the Digital Imaging and Communications in Medicine (DICOM) dataset acquired from MR, CT, and cone beam CT (Figure 1A), into a Stereolithography (STL) file format (Figure 1B) native to the stereolithography Computer Aided Design (CAD) software created by 3D systems. In essence, this means data from patient-specific anatomy can be (relatively) easily converted to a language computers and three-dimensional printers understand. From this data, an anatomical model of patient-specific anatomy (Figure 1C) can be created in less than a day. Previously this process would require weeks or months. In the medical realm, the most advanced application of this process is in the recreation of patient-specific anatomy of cerebral, cardiac, and peripheral vasculature. After a three-dimensional model is created, a replica of the vasculature can be recreated in silicone via various methods—dip coating or the core shell methods. Dip coating involves coating the outer wall of the three-dimensional model, then eventually dissolving away the core itself, resulting in a silicone model of the vessel (Figure 1D). The core shell method creates a cast similar to investment casting, where the silicone can be poured into the cast, and when the cast is removed, the result is a silicone structure or model based on the initial imaging that was acquired. Again, the advantage is that unique patient-specific anatomy can now be created in a very short timeline compared to prior methods.

Now that patient-specific anatomy can be physically, not virtually, recreated, it provides a platform for replication of procedures that virtual simulation cannot recreate. As the friction coefficient of silicone is orders of magnitude higher than the inner wall of blood vessels, a coating to reduce that friction coefficient is required so that catheters, wires, and implantable devices behave similarly to the clinical conditions. It is now possible to recreate intracranial aneurysms based on patient-specific anatomy and perform an endovascular procedure in a silicone model that is indistinguishable from the clinical procedure itself (Figure 2). Currently, practicing on complex aneurysms is possible prior to the actual procedure itself. This innovative technology was highlighted in a live demo session during the 2016 CNS Annual Meeting.

Figure 2. The Vascular Simulations Replicator. A replica of the human arterial tree and left side of the human heart. The vasculature can also be customized to be patient-specific based on clinical radiologic imaging such as CT, MRI or cone beam CT.

Following, the physical model is agnostic to the development of new technology because there is no interpretation as to how that novel device will perform when the practitioner is performing the procedure with the actual device. This has tremendous implications, not just for the resident or fellow in training, but for experienced interventionalists, who can now gain insight into the behavior of devices prior to performing the procedure on the patient in a scenario that is identical to what he or she will eventually encounter. Furthermore, the interventionalist can recreate those failure modes, such as herniation of coils, into the parent vessel that now require removal of coils, a scenario that the interventionalist would prefer never to occur clinically. This is akin to Sully landing in the Hudson River! It is a scenario you want to avoid at all costs, but if it occurs, you are ready and familiar with the maneuvers to bail out of that situation without the cost of human life or disability.

In medicine and certainly neurosurgery, good judgment comes from experience, and experience comes from bad clinical judgment or a mistake that has occurred by yourself or someone else. In the past, this frequently resulted in collateral damage to the patient. The era of medical simulation, despite recent advances, is still in its infancy. Medical training is evolving rapidly. There is no doubt that medical simulation will be critical to helping practitioners abide by the Hippocratic oath we all took to “First do no harm.”


Sampath SA, Voon SH, Davies H. Factors affecting the learning curve in Computer Assisted Total Knee Arthroplasty. Conf Proc IEEE Eng Med Biol Soc. 2008; 2008:3239-40.

Gladwell M. Outliers: the story of success. 2008. New York: Little, Brown and Co.

Ericsson KA, Krampe RT, Tesch-Romer C. The Role of Deliberate Practice in the Acquisition of Expert Performance. Psychological Review: 100 (3) 363-406

Kunkler K. The role of medical simulation: an overview. The International Journal of Medical Robotics + Computer Assisted Surgery: MRCAS. 2006; 2(3):203-10.

Gross BC, Erkal JL, Lockwood SY, et al. Evaluation of 3D printing and its potential impact on biotechnology and the chemical sciences. Anal Chem. 2014; 86(7):3240-53.

Disclosures: The authors have an interest in Vascular Simulations, LLC.